Model Taxonomy · Model Landscape

Comprehensive guide to the model landscape as of October 1, 2026, covering model families, capabilities, and selection criteria for production systems.

This chapter provides a comprehensive guide to the model landscape as of October 1, 2026, covering model families, capabilities, and selection criteria for production systems.

Last verified: October 1, 2026. The model landscape evolves rapidly. Always cross-check with provider pricing pages and release notes.

October 2026 headline (mid-August to October 1): Anthropic, OpenAI, Google, SpaceXAI, and Meta all shipped new models in September, and the story is price as much as capability. Anthropic shipped Claude Fable 5.1 and Mythos 5.1 (September 1, same $10/$ 50 with cache reads cut to 0.025x), Claude Opus 5.5 (September 22, the first Opus price cut since Opus 4.5, to $4/$ 20, and now Anthropic's "start here" default), and Claude Sonnet 5.5 (September 28, $2/$ 10). OpenAI shipped GPT-6 Astra (September 3, $10/$ 50, a Daybreak-first staged rollout and the first OpenAI model rated Critical for cybersecurity), GPT-6 Sol and Luna (September 22, $2/$ 10 and $0.10/$ 0.50, about half the GPT-5.6 prices they replace, with no GPT-6 Terra), and GPT-6.1 Sol (September 29, same $2/$ 10 with 0.05x cache reads). Google shipped Gemini 3.8 Flash (September 2, its third Flash in six weeks, half price until January 1, 2027) plus a Fairwind-gated 3.8 Flash Cyber, then on September 30 announced Gemini 4 Argon, which is rolling out to cyber defenders only and has no API model ID yet. SpaceXAI shipped Grok 4.7 (September 21) on a new, larger base at an unchanged $2/$ 6, and Meta shipped Muse Spark 1.3 (September 2). The practical consequences: the mid tier converged on $2/$ 10 (Sonnet 5.5, GPT-6 Sol, GPT-6.1 Sol), cache-read multipliers now differ by model (0.1x, 0.05x, 0.025x), OpenAI bills the whole request at long-context rates above 272K input, and the newest Claude models reject forced tool_choice and bind thinking blocks to the producing model and conversation, which breaks naive routers. Benchmarks reset too: Artificial Analysis re-baselined its Intelligence Index (v4.3, September 7; earlier scores are not comparable), Terminal-Bench 4.0 replaced 3.0, the public SWE-Bench Pro split saturated, and Claude scores now carry a "with fallback" label. On open weights, Z.ai shipped GLM-5.3's weights on schedule (August 28) under an MIT-plus license, Xiaomi's MiMo-V2.6-Pro (September 21, MIT) took the top open-weight spot, DeepSeek V4.1-Flash (September 10, MIT) cut Flash prices less than a month after the August 16 increase, and Qwen's new Community License requires a separate license for any MaaS or coding/office-assistant business. Retirements: the Assistants API shut down August 26, gpt-5.4-cyber was shut down October 1 on 20 days' notice, the original gpt-4o-2024-05-13, o1, o3-mini, and o4-mini snapshots leave the OpenAI API October 23, and Claude Sonnet 4.5 retires November 30. Vendor-reported numbers below are labeled as such.

August 2026 headline: Most of the month's releases were post-training refreshes or derivatives rather than new pretraining runs, and no US frontier lab shipped a new flagship base model. The action moved to prices, licenses, and access control. The clear exception is Alibaba's Qwen3.8-Max, a genuinely new 2.4T base whose weights landed August 12. Claude Sonnet 5's introductory $2/$ 10 per 1M became the permanent price on August 10, and the scheduled September 1 rise to $3/$ 15 was canceled, making Sonnet 5 permanently cheaper than the Sonnet 4.6 it replaced. Running the other way, DeepSeek raised V4 prices 3x to 12x effective August 16 at 16:00 UTC, moving to peak and off-peak billing (off-peak is exactly half peak, and peak covers only 01:00-04:00 and 06:00-10:00 UTC on weekdays, excluding Chinese public holidays). That ended its run as the unambiguous cheap option. OpenAI shipped GPT-5.6-Cyber (August 10, $12.50/$ 75 per 1M) behind a new two-tier Daybreak Blue and Daybreak Red access program, the most concrete production example yet of capability-tiered gating. Google released Gemini 3.7 Flash (August 13) at half price through year-end, and SpaceXAI released Grok 4.6 (August 12) with a 500K context. On open weights the licensing picture split: Alibaba's Qwen3.8-Max (August 12, 2.4T/95B active) shipped under a bespoke commercially gated license while its Qwen3.8-27B sibling (August 14) shipped plain Apache 2.0, Meta returned to open weights with Muse Glimmer (August 10, 30B, Apache 2.0), and Z.ai withheld GLM-5.3's weights pending safety evaluation after cyber capability grew faster than expected (the weights then shipped on schedule, August 28). Claude Opus 4.1 retired on August 5, closing out the last $15/$ 75 Opus tier. Benchmark figures throughout this section are largely vendor-reported; confirm on independent leaderboards.

July 2026 headline: Anthropic shipped a full generation refresh: Claude Sonnet 5 (June 30, claude-sonnet-5, new default everywhere, introductory $2/$ 10 per 1M, made permanent on August 10) and Claude Opus 5 (July 24, claude-opus-5, unchanged $5/$ 25 with an optional Fast mode at $10/$ 50 about 2.5x faster). Claude Fable 5 was restored globally July 1 after the export-control suspension, with a new jailbreak-specific cybersecurity classifier; Mythos 5 returned only to roughly 100 US critical-infrastructure organizations via Project Glasswing. GPT-5.6 (Sol, Terra, Luna) reached general availability July 9, and on July 30 OpenAI cut Luna 80% to $0.20/$ 1.20 and Terra 20% to $2/$ 12 (Sol stayed at $5/$ 30 until a promotional cut to $4/$ 20 on August 21). The open-weight frontier had its strongest month ever: Moonshot Kimi K3 (July 16, weights July 27) is the largest open-weight model to date at 2.8T total / 104B active with a 1M context, and Thinking Machines Lab debuted Inkling (July 15, 975B / 41B active, open weights), the leading US open-weights model at the time. Google shipped Gemini 3.6 Flash, 3.5 Flash-Lite, and a government-gated 3.5 Flash Cyber (July 21) while delaying Gemini 3.5 Pro. Meta Muse Spark 1.1 (July 9) arrived with Meta's first paid self-serve model API ( $1.25/$ 4.25 per 1M). Black Forest Labs announced FLUX 3 (July 23), a unified image, video, audio, and action model, in gated early access. Benchmark figures across these launches are largely vendor-reported; confirm on independent leaderboards.

June 2026 headline: Anthropic released Claude Fable 5 (June 9, claude-fable-5, $10/$ 50 per 1M, 1M context), its most capable widely released model: a Mythos-class model made safe for general availability, with an Opus 4.8 fallback safeguard on sensitive topics. Claude Mythos 5 ships the same day as the unrestricted variant for Project Glasswing partners, succeeding Mythos Preview at less than half its price.

June 10-26 update: A dense second wave of June launches followed. Google DeepMind DiffusionGemma (June 10, Apache 2.0) is Google DeepMind's first open-weight text-diffusion model: a 26B Mixture-of-Experts (~4B active) that denoises blocks of tokens in parallel for roughly 4x faster generation on a single H100, trading some quality versus standard Gemma 4. Gemini 3.5 Live Translate (June 9 per launch coverage; Google's changelog has no entry for it, and the current ID is gemini-3.5-live-translate-preview) added real-time speech-to-speech translation across 70+ languages in public preview via the Gemini Live API and AI Studio. Cohere North Mini Code 1.0 (June 9, Apache 2.0) is Cohere's first open coding model, a 30B / 3B-active MoE that runs on one H100. Moonshot Kimi K2.7 Code (June 12, Modified MIT) tunes K2.6 for long-horizon software work (1T / 32B-active MoE, roughly 30% fewer thinking tokens). Z.ai GLM-5.2 (coding-plan access June 13, open weights under MIT June 16-17) is a 744B / 40B-active MoE with a 1M context that reports SWE-Bench Pro 62.1, ahead of GPT-5.5 on that benchmark, at roughly $1.40 /$ 4.40 per 1M. xAI Grok Imagine Video 1.5 reached general availability June 16 (image-to-video with synchronized audio, $0.080 per second of video), and **Grok 4.3** arrived on Amazon Bedrock June 15 ($ 1.25 / $2.50 per 1M, xAI's first model there). Alibaba's official Qwen Cloud changelog lists a June snapshot adding vision to Qwen 3.7-Max (text-only at its May launch), though some independent coverage attributes that vision update to Qwen 3.7-Plus instead, so verify before relying on it. Separately, on June 12 Anthropic suspended access to Claude Fable 5 and Claude Mythos 5 following a US export-control directive, with Mythos 5 later cleared for a limited set of US institutions. Then on June 26, OpenAI previewed GPT-5.6 (Sol, Terra, and Luna), its next-generation line, in a limited release to a small set of US-government-approved partners over dual-use cybersecurity concerns, echoing the Anthropic restriction; Sol claims a new Terminal-Bench 2.1 record and Terra targets GPT-5.5-level quality at about half the cost. Coding scores here are largely vendor-reported; confirm on independent leaderboards.

May 2026 recap: Anthropic Claude Opus 4.8 (May 28, same $5/$ 25 price as Opus 4.7; Dynamic Workflows research preview with hundreds of parallel subagents; fast mode at $10/$ 50 is 3x cheaper than the Opus 4.7 fast mode); OpenAI GPT-5.5 (April 23) and GPT-5.5 Instant (May 5, default in ChatGPT); Claude Opus 4.7 (April 16, GA on Bedrock/Vertex/Foundry); Google Gemma 4 (April 2, Apache 2.0) and Gemini 3.2 Flash (quiet rollout May 5); DeepSeek V4 Pro and V4 Flash (April 24; 75% V4 Pro discount made permanent May 22, new list price $0.435/$ 0.87 per 1M from June 1); Moonshot Kimi K2.6 (April 20, 1T MoE / 32B active); Alibaba Qwen 3.6 Plus / 3.6-35B-A3B / 3.6 Max-Preview; Mistral Medium 3.5 (April 29, unified chat/reasoning/coding/vision); Meta Muse Spark (April 8, first closed-weight Meta model); Llama 4 Behemoth release paused through fall 2026 amid capability concerns. SWE-bench Verified published leaders before the Fable 5 launch: Claude Mythos Preview 93.9%, GPT-5.5 88.7%, Claude Opus 4.8 88.6%; ARC-AGI-2 leader: GPT-5.5 at 85.0%. Anthropic describes Fable 5 as state of the art on nearly all tested benchmarks; standard numeric scores were not in the launch post, so verify on the leaderboards.

Model Categories
Frontier Models (October 2026)
Reasoning Models
Open Source Models
Specialized Models
Embedding Models
Model Selection Framework & Semantic Routing
Sovereign AI & Data Residency
Capability Comparison
Interview Questions
References

Model Categories

By Capability Level (October 2026)

Tier	Characteristics	Examples	Use Case
Frontier	Highest capability ceiling; $10/$ 50 class or gated	Claude Fable 5.1, GPT-6 Astra, Claude Opus 5.5 (at $4/$ 20); Gemini 4 Argon announced, not GA	Hardest reasoning, long-horizon agents, ceiling-bound coding
Workhorse ( $2/$ 10 tier)	Near-frontier agentic quality at mid-tier price	Claude Sonnet 5.5, GPT-6 Sol, GPT-6.1 Sol, Grok 4.7 ( $2/$ 6), Muse Spark 1.3 ( $1.25/$ 4.25)	Production agents, coding at scale, the default routing tier
Fast/Efficient	Lowest cost per token, high volume	GPT-6 Luna, Gemini 3.8 Flash, Claude Haiku 4.5, DeepSeek V4.1-Flash	Classification, extraction, routing, high-volume streaming
Previous Generation (still served)	Well characterized, often superseded within weeks	Claude Opus 5 / Sonnet 5 / Fable 5, GPT-5.6 Sol / Terra / Luna, Claude Opus 4.8, Gemini 3.7 Flash	Workloads that value stable, already-evaluated behavior
Small/Edge	Private, edge, single GPU	Gemma 4 (12B, E4B), Qwen3.8-27B, IBM Granite 4.2, Muse Glimmer 30B	Local privacy, on-device, single-GPU serving
Reasoning-Heavy	Extended internal reasoning at high effort	GPT-6 Astra (xhigh/max), Claude Fable 5.1, Claude Opus 5.5 at high effort, DeepSeek V4.1-Flash (effort 1-100)	Math, code debugging, multi-step logic

Anthropic's only Haiku is still Haiku 4.5. Claude Haiku 5.5 was announced for "the coming weeks" but had not shipped as of October 1.

By Reasoning Mode (October 2026)

Mode	Capability	Models	Use Case
Minimal reasoning	The fastest path the model allows	GPT-6.1 Sol at `low`, GPT-6 Luna at its lowest effort, Gemini 3.8 Flash at `low`, Claude Sonnet 5.5 at `between_tools`, Claude Haiku 4.5	Chat, simple extraction
Always-on reasoning	Thinking cannot be switched off	Claude Opus 5.5 and Fable 5.1 (adaptive thinking always on), GPT-6 Astra (no `none` effort; `low` to `max`)	Math, code debugging, planning
Dial-controlled	Effort or thinking level per request	GPT-6.1 Sol (`low` to `max`, default `medium`), Claude Opus 5.5 (default `medium`), Gemini 3.8 Flash (`thinking_level`), DeepSeek V4.1-Flash (1-100)	Variable-complexity tasks

The off switch is going away. Sonnet 5.5 returns a 400 on thinking: {type: "disabled"}, Opus 5.5 cannot disable thinking, and Astra has no none effort. Sampling knobs are going too: the Claude API returns a 400 for non-default temperature and top_p on Opus 4.7 and later, Astra accepts no temperature, top_p, or logprobs, and Google deprecated those parameters in the Gemini API on July 21. Effort is now the control surface, and defaults move between versions (Opus 5 defaulted to high, Opus 5.5 to medium), so set it explicitly on every route.

Frontier Models (October 2026)

Claude Opus 5.5 (Anthropic) - September 2026 NEW

Attribute	Value
Model ID	`claude-opus-5-5` (Bedrock: `anthropic.claude-opus-5-5`)
Context Window	1M tokens; 128K max output (300K on Batch with the beta header)
Input / Output Cost	$4.00 /$ 20.00 per 1M (Opus 5: $5 /$ 25)
Cache / Batch	Cache write $5.00 (5 min) or$ 8.00 (1 hr); cache read $0.20 (0.05x input); Batch$ 2 / $10; minimum cacheable prompt 512 tokens
Fast mode	$8 /$ 40 per 1M (research preview, Claude API only), down from $10 /$ 50 on Opus 5 and Opus 4.8
Thinking	Adaptive thinking always on (cannot be disabled); default effort `medium` (Opus 5 defaulted to `high`)
Availability	Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, Claude Platform on AWS; zero data retention available
Benchmarks	Anthropic-reported: Terminal-Bench 4.0 66.4% at xhigh (SE ±2.6) vs Fable 5.1 55.8% (Anthropic's run) and GPT-6 Astra 57.9% (OpenAI's figure, high effort); OSWorld 2.1 partial 81.8%; GDPval-AA v2.1 1846 Elo; HLE with tools 67.7%. Independent: Artificial Analysis Intelligence Index v4.3.2 58 (max, with fallback), first overall
Released	September 22, 2026 (knowledge cutoff June 2026; retirement not sooner than September 22, 2027)

What it is: The second Opus price cut (after Opus 4.5 came down from $15/$ 75), and the model Anthropic's docs now tell you to start with for most workloads. Fable 5.1 is reserved for demanding reasoning, long-horizon agentic work, or cases where Opus 5.5 at higher effort still falls short in your evals. A cheaper model that beats the more expensive Fable 5.1 on most of Anthropic's own agentic benchmarks inverts the usual "pay more for the ceiling" routing logic. Anthropic claims about 40% lower cost than Opus 5 on typical workloads (vendor-reported).

Migration traps from Opus 5: the default effort dropped from high to medium, so an unpinned upgrade changes both latency and quality; forced tool_choice (any or tool) returns a 400; thinking blocks are bound to the producing model and conversation; and on the Claude API and Google Cloud only computer_toolset_20260801 is accepted for computer use. Opus 5.5 adds a biology safety classifier alongside the cyber one. When safeguards intervened during Anthropic's evals, cyber tasks were completed by Opus 4.8 and biology and frontier-LLM-development tasks by Opus 5, so the headline scores belong to a routed system, not one set of weights.

Claude Sonnet 5.5 (Anthropic) - September 2026 NEW

Attribute	Value
Model ID	`claude-sonnet-5-5`
Context Window	1M tokens; 128K max output
Input / Output Cost	$2.00 /$ 10.00 per 1M (unchanged from Sonnet 5)
Cache / Batch	Cache write $2.50 (5 min) or$ 4.00 (1 hr); cache read $0.20; Batch$ 1 / $5
Thinking	Adaptive thinking on by default; the lowest setting is `between_tools` (accepted at `high` effort or below), and `thinking: {type: "disabled"}` returns a 400. Default effort `high` on the API, medium in Claude Code and the Claude apps
Availability	Claude API, Bedrock, Google Cloud, Foundry, Claude Platform on AWS; zero data retention available
Benchmarks	Anthropic-reported, with Opus 5.5 in parentheses: Terminal-Bench 4.0 70.6% (66.4%), CursorBench 4.0 55.5% (57.8%), GDPval-AA v2.1 1844 (1846), OSWorld 2.1 partial 80.1% (81.8%), HLE with tools 64.5% (67.7%), FrontierCode 1.1 46.2% at max effort (54.4%); Anthropic notes Sonnet 5.5 scored lower at max than at xhigh on this test. Independent: AA Intelligence Index v4.3.2 56 (max, with fallback)
Released	September 28, 2026 (knowledge cutoff June 2026; retirement not sooner than September 28, 2027)

What it is: If these numbers hold up independently, a $2/$ 10 model sits within a few points of the flagship on most of Anthropic's own vendor-reported agentic evals (about 8 points behind on FrontierCode 1.1), which collapses the case for sending most agent traffic to Opus. Anthropic claims 30%+ faster output than Sonnet 5 and up to 30% lower cost per task (vendor-reported). Anthropic's Sonnet 5.5 table puts Sonnet 5 at 10.3% on Terminal-Bench 4.0, while its Opus 5.5 table puts Opus 5 at 52.3%; treat the Sonnet 5 figure as an outlier and do not use it as a baseline.

Breaking changes vs Sonnet 5: disabling thinking returns a 400 (send between_tools instead); forced tool use returns a 400; thinking blocks are bound to the model, the conversation, and the producing account (blocks sent from an unlinked account are dropped and recorded as organization_binding_mismatch); computer_20251124 is rejected on the Claude API and Google Cloud; and non-default temperature, top_p, and top_k still return a 400. Sonnet 5 became a legacy model three months after launch.

Claude Fable 5.1 and Mythos 5.1 (Anthropic) - September 2026 NEW

Attribute	Value
Model IDs	`claude-fable-5-1` (GA; Bedrock `anthropic.claude-fable-5-1`); `claude-mythos-5-1` (Project Glasswing only, per the docs)
Relationship	Same weights, different safeguards; Mythos 5.1 is the more permissive variant
Input / Output Cost	$10.00 /$ 50.00 per 1M (unchanged from Fable 5)
Cache / Batch	Cache write $12.50 (5 min) or$ 20.00 (1 hr); cache read $0.25 (0.025x input, down from$ 1.00 on Fable 5); Batch $5 /$ 25
Context Window	1M tokens at standard pricing; 128K max output
Thinking	Adaptive thinking always on; default effort `high`
Data handling	30-day retention required; not available under zero data retention unless Anthropic expressly authorizes it (customers eligible for Enterprise Frontier Safeguards can use Fable 5.1 and Fable 5 with ZDR until EFS is ready)
Provenance	Anthropic's statistical text watermark on outputs on every platform; signed C2PA Content Credentials on generated media retrieved via the Files API
Benchmarks	Terminal-Bench 4.0: 57.88% at max on the tbench.ai leaderboard, 55.8% in Anthropic's own run. AA Intelligence Index v4.3.2: 53 (max, with fallback)
Released	September 1, 2026 (knowledge cutoff June 2026; retirement not sooner than September 1, 2027)

What changed: the cost of staying on the ceiling model for long agent loops. Anthropic says the 75% cache-read cut lowers typical bills about 25% and highly agentic ones up to 45%, and that its newest cyber safeguards produce about 60% fewer false positives than Fable 5's earlier ones (both vendor-reported). Fable 5.1 may now be used to discover vulnerabilities but not to develop exploits. Fable 5 remains available.

Mythos 5.1 access: Glasswing participants only, as of October 1. Anthropic says it will also be offered through the Life Sciences Verification Program (first participants enrolled with the US government) and, "in the near future", through the existing Cyber Verification Program, which today covers Opus- and Sonnet-class models. For now Anthropic says Mythos 5.1 is available only to a set of US organizations.

Safeguard fallback is now an API contract. On Fable 5 a classifier hit handed the request to Opus 4.8 automatically. On Fable 5.1 it is explicit and configurable:

Diagram

From September 24, refusals that arrive before any output in the bio, frontier_llm, and reasoning_extraction categories are billed on all platforms (other categories stay unbilled); mid-stream refusals bill the input plus the output already streamed; and fallback credit refunds the prompt-cache cost of switching models. Budget for refusals, and route on stop_reason, not HTTP status.

Claude 5.5 / 5.1 Migration Checklist

Change	Applies to	What breaks	Do this instead
Forced `tool_choice` (`any` or `tool`) returns 400	Fable 5.1, Mythos 5.1, Opus 5.5, Sonnet 5.5	Structured extraction that forces a tool call	`auto` plus strict tool use, or structured outputs
Thinking blocks bound to model and conversation	Fable 5.1, Opus 5.5, Sonnet 5.5 (Mythos 5.1 skips the prefix check)	Replaying history to another model silently drops reasoning; editing anything before a thinking block returns 400 for accounts created on or after August 31, 2026	Treat history as append-only; opt into `prefix_mismatch_behavior: "drop_block"` with the `thinking-binding-controls-2026-08-01` beta header
Thinking cannot be disabled	Opus 5.5; Sonnet 5.5's floor is `between_tools`	Latency budgets that assumed thinking off	Set effort explicitly and re-measure latency
Default effort changed	Opus 5.5 is `medium` (Opus 5 was `high`)	Silent quality and latency shift on upgrade	Pin effort on every request
`computer_20251124` rejected	Opus 5.5 and Sonnet 5.5 on the Claude API and Google Cloud (Bedrock still accepts it)	Computer-use agents on the old tool version	Move to `computer_toolset_20260801`
Pre-output refusals billed in three categories	All platforms since September 24	Cost models that treat refusals as free	Track `stop_reason: "refusal"` as a cost line

Two smaller changes trip harnesses: text between tool calls now arrives in thinking blocks that are empty at the default display: "omitted" (the display: "updates" beta returns the text), and supporting betas shipped for per-message effort and turn-scoped system messages (September 1), compaction on demand (compact-2026-09-04, September 14), and inline tool definitions mid-conversation without cache loss (inline-tools-2026-09-15, September 22).

Claude Haiku 4.5 (Anthropic) - Still the Only Haiku

Haiku 4.5 remains Anthropic's fast tier. Its retirement floor is "not sooner than October 15, 2026", but it had no deprecation notice as of October 1, and with Anthropic's 60-day minimum notice it should not retire on the Claude API before about December. Microsoft Foundry lists claude-haiku-4-5 for retirement on November 15, 2026. Haiku 5.5 is announced, not shipped, so fast-tier designs built on Haiku 4.5 should keep a GPT-6 Luna or Gemini 3.8 Flash fallback ready and plan an eval pass when Haiku 5.5 lands.

Claude Opus 5 (Anthropic) - July 2026, SUPERSEDED by Opus 5.5

Attribute	Value
Model ID	`claude-opus-5`
Context Window	1M tokens (default and max; 128K max output)
Input / Output Cost	$5.00 /$ 25.00 per 1M (unchanged from Opus 4.8)
Fast mode	$10.00 /$ 50.00 per 1M, about 2.5x faster
Benchmarks	Anthropic's launch post reported the pre-release benchmark it called Frontier-Bench v0.1 (later Terminal-Bench 3.0, itself replaced by Terminal-Bench 4.0 on August 28; scores do not carry over): 43.3% at max effort vs GPT-5.6 Sol's 34.4%, Fable 5's 33.7%, and Opus 4.8's 18.7%; within 0.5% of Fable 5 on CursorBench 3.2 at about half the cost per task. Vendor-reported. Since then: Terminal-Bench 4.0 leaderboard 53.94% at xhigh; SWE-Bench Pro v2 (Scale, September 22) 99.4% on the saturated public split and 81.6% (222/272) on the private set, the top private score Scale published
Released	July 24, 2026 (Claude API, Claude Code, Claude Cowork; new default on Claude Max)

What it is: The Opus line's generational successor at unchanged pricing, aimed at long-horizon agentic coding and computer use. Beta features include mid-conversation tool changes and automatic fallback routing. The dual-price Fast mode continues the pattern Opus 4.8 introduced: one model, two latency tiers.

Status: Superseded by Opus 5.5 on September 22 at a lower price ( $4/$ 20 vs $5/$ 25). Still available as a legacy model, and one of Fable 5.1's two permitted safeguard fallback targets (with Opus 4.8).

Claude Sonnet 5 (Anthropic) - June 2026, SUPERSEDED by Sonnet 5.5

Attribute	Value
Model ID	`claude-sonnet-5`
Context Window	1M tokens per third-party coverage (not stated in the launch post)
Input / Output Cost	$2.00 /$ 10.00 per 1M (permanent since August 10, 2026)
Cache / Batch	Cache write $2.50 per 1M (5 min) or$ 4.00 (1 hr); cache hit $0.20; Batch API$ 1.00 / $5.00
Positioning	The most agentic Sonnet at launch: planning, browser and terminal tool use, autonomous operation approaching Opus 4.8 at lower cost
Safety posture	Cyber safeguards on by default; deliberately reduced cybersecurity capability relative to Opus-class models
Released	June 30, 2026 (default model across consumer and developer products same day)

What it is: The production workhorse that replaced Sonnet 4.6 as the default. On August 10, 2026 Anthropic made the introductory $2/$ 10 rate permanent and canceled the September 1 increase to $3/$ 15, so Sonnet 5 is permanently cheaper than the Sonnet 4.6 it succeeded (still $3/$ 15).

Status: Superseded by Sonnet 5.5 on September 28 at the same price; still available as legacy. Separately, Anthropic deprecated Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) on September 30, with retirement on November 30, 2026 on the Claude API and Foundry and claude-sonnet-5-5 as the named replacement.

Claude Fable 5 (Anthropic) - June 2026, SUPERSEDED by Fable 5.1

Attribute	Value
Model ID	`claude-fable-5`
Context Window	1M tokens (Opus 4.7 tokenizer; roughly 30% more tokens than pre-4.7 models for the same text)
Max Output	128K tokens
Input Cost	$10.00 / 1M tokens
Output Cost	$50.00 / 1M tokens
Thinking	Adaptive thinking, always on (no separate extended-thinking toggle)
Multimodal	Text + Vision (new state of the art on vision tasks per Anthropic)
Benchmarks	State of the art on nearly all tested benchmarks per Anthropic; highest frontier score on Cognition's FrontierCode, highest on the Hebbia Finance Benchmark, ViBench, and CursorBench. Standard numeric scores (SWE-bench, GPQA) were not published in the launch post.
Released	June 9, 2026 (GA on Claude API, Claude Platform on AWS, Amazon Bedrock, Vertex AI, Microsoft Foundry)

What it is: A Mythos-class model made safe for general availability. Until then the Mythos line (SWE-bench Verified 93.9% on Mythos Preview) was restricted to ~11 Project Glasswing partners over dual-use cybersecurity concerns. Fable 5 brought that capability tier to everyone by pairing it with conservative safeguards.

The Opus 4.8 fallback safeguard: When Fable 5's classifiers detected a request in one of three categories (offensive cyber techniques, bioweapon-adjacent biology and chemistry, or attempts to distill the model), the response was delegated to Claude Opus 4.8 and the user was informed. Anthropic said this triggered in under 5% of sessions and was deliberately tuned conservative, so some harmless requests got caught. Architecturally this is a production example of model-tier routing as a safety control, not just a cost control. Fable 5.1 turned it into the explicit, billable API contract described above.

Considerations: 2x the per-token price of Opus 4.8 ( $10/$ 50 vs $5/$ 25). Mythos-class traffic carries a 30-day data retention requirement (not used for training; access-logged; deleted after 30 days in almost all cases), which matters for compliance reviews. On subscription plans it was included at no extra cost June 9-22, then moved to usage credits. Fable 5 cache reads were $1.00 per 1M; Fable 5.1 cut them to$ 0.25 at the same list price, so there is little reason to start new work on Fable 5.

Claude Mythos 5 (Anthropic) - June 2026, SUPERSEDED by Mythos 5.1

Attribute	Value
Model ID	`claude-mythos-5`
Status	Limited availability: Project Glasswing partners and select biology researchers
Relationship	Same underlying model as Fable 5 with safeguards lifted in some areas
Pricing	$10 /$ 50 per 1M (less than half of Mythos Preview)
Released	June 9, 2026

Why it matters: Succeeded Claude Mythos Preview at comparable or somewhat stronger capability and much lower price. The Fable/Mythos split formalized a two-track release pattern: one safeguarded general release, one unrestricted release for vetted defenders. Mythos 5.1 continues it.

Claude Opus 4.8 (Anthropic) - May 2026, Legacy

Attribute	Value
Context Window	1M tokens (standard pricing across the full window)
Input Cost	$5.00 / 1M tokens (same as 4.7)
Output Cost	$25.00 / 1M tokens (same as 4.7)
Cache: 5m write	$6.25 / 1M tokens
Cache: 1h write	$10.00 / 1M tokens
Cache: hit / refresh	$0.50 / 1M tokens
Batch API	$2.50 /$ 12.50 per 1M (50% discount)
Fast mode (research preview)	$10 /$ 50 per 1M (about 2.5x faster; 3x cheaper than the Opus 4.7 fast mode which was $30 /$ 150)
Extended Thinking	Native, adaptive mode
Multimodal	Text + Higher-resolution Vision
SWE-bench Verified	88.6%
SWE-Bench Pro	69.2% (up from 64.3% on Opus 4.7)
Terminal-Bench 2.1	74.6% (GPT-5.5 led at 78.2% at release)
GDPval-AA	1890 Elo (up from 1753 on Opus 4.7)
OSWorld-Verified	82.3%
Online-Mind2Web	84%
Released	May 28, 2026 (GA on Claude API, AWS Bedrock, Vertex AI)

Best for: Long-running autonomous coding work in Claude Code, codebase-scale migrations, agentic workflows that need parallel subagents, and workloads where the alignment and honesty gains matter. It is also one of Fable 5.1's two permitted safeguard fallback targets (with Opus 5), and Anthropic's Opus 5.5 evals routed cyber tasks to it when safeguards intervened, so it stays in production paths even for teams that never call it directly.

Key features over Opus 4.7:

Dynamic Workflows (research preview): Claude plans the work and runs hundreds of parallel subagents in a single Claude Code session, verifies their outputs, and reports back. Suited for codebase-scale migrations across hundreds of thousands of lines.
Mid-task system messages: The Messages API now accepts system messages mid-conversation, useful for steering long agent runs without ending the session.
Optional fast mode at roughly 2.5x speed for $10 /$ 50 per 1M, priced 3x lower than the Opus 4.7 fast mode.
Effort-control toggle in claude.ai and Cowork lets users tune reasoning depth per turn.
Expanded Claude Code rate limits.

Considerations: Tokenizer is the same one introduced in Opus 4.7: roughly 30% more tokens than the pre-4.7 tokenizer for the same text, per Anthropic's token-counting docs, with the exact increase depending on content (the Opus 4.7 launch post gave a range of 1.0x to 1.35x). At release, GPT-5.5 held the SWE-Bench Verified leaderboard at 88.7% and led Terminal-Bench 2.1 at 78.2%. GPQA Diamond slipped 0.6 pts versus Opus 4.7. Anthropic's tokenizer change means token counts and bills for the same text are not directly comparable to pre-4.7 models. There was no Claude Sonnet 4.8; the line jumped to Claude Sonnet 5 on June 30, 2026, which replaced Sonnet 4.6 as the production workhorse. For new work, Opus 5.5 is cheaper ( $4/$ 20) and stronger.

NOTE
Retired August 5, 2026: claude-opus-4-1-20250805 was removed from the Claude API, closing out the last $15/$ 75 per 1M Opus tier. At that point every first-party Opus SKU was $5/$ 25 ( $10/$ 50 in Fast mode); Opus 5.5 later cut the current Opus to $4/$ 20. Amazon Bedrock and Google Cloud set their own schedules: on Bedrock, Opus 4.1 moves into higher-priced public extended access on October 8, 2026 and reaches end of life on January 8, 2027. Code pinned to that ID fails on the first-party API while still working on the partner clouds.

Claude Opus 4.7 (Anthropic) - April 2026, Legacy

Attribute	Value
Context Window	1M tokens
Max Output	128K tokens
Input Cost	$5.00 / 1M tokens (same as 4.6)
Output Cost	$25.00 / 1M tokens (same as 4.6)
Extended Thinking	Native, Adaptive mode
Multimodal	Text + Higher-resolution Vision
SWE-bench Verified (Adaptive)	87.6% (May 13, 2026)
Released	April 16, 2026 (GA on API, Bedrock, Vertex, Microsoft Foundry)

Status: Still available as a legacy model. It powered Claude Code at release; Claude Code's default Opus has been Opus 5.5 since 2.1.280. Prefer Opus 5.5 for new work: it is cheaper and stronger.

Claude Mythos Preview (Anthropic) - SUCCEEDED BY MYTHOS 5 AND 5.1

Attribute	Value
Status	Restricted research preview, Project Glasswing partners only (~11 orgs: AWS, Apple, Cisco, Google, Microsoft, NVIDIA, Palo Alto, etc.); deprecated since June 9, 2026, retirement date to be announced
Reason for restriction	Dual-use cybersecurity capabilities
SWE-bench Verified	93.9% (May 13, 2026; the published SOTA before the Fable 5 / Mythos 5 launch)
Released	April 7, 2026 (restricted partner preview); succeeded by Claude Mythos 5 on June 9, 2026 at less than half the price

Best for: Historical reference. Its capability tier reached general availability as Claude Fable 5 on June 9, 2026; new Glasswing work should target Mythos 5.1.

Claude Opus 4.6 (Anthropic) - February 2026, Legacy

Attribute	Value
Context Window	1M tokens
Max Output	128K tokens
Input Cost	$5.00 / 1M tokens
Output Cost	$25.00 / 1M tokens
Extended Thinking	Native adaptive thinking (configurable budget_tokens)
Multimodal	Text + Vision
Highlights	Anthropic's flagship at release; now legacy
Released	February 2026

Status: Still available. Opus 5.5 is cheaper ( $4/$ 20 vs $5/$ 25) and stronger, so there is no cost reason to stay on it.

Claude Sonnet 4.6 (Anthropic) - February 2026, Legacy

Attribute	Value
Context Window	1M tokens
Input Cost	$3.00 / 1M tokens
Output Cost	$15.00 / 1M tokens
Extended Thinking	Supported
Multimodal	Text + Vision
Highlights	Handled tasks that previously required the Opus tier; the cost/quality default through mid-2026
Released	February 2026

Status: Still available, but at $3/$ 15 it now costs more than Sonnet 5.5 ( $2/$ 10). Migrate unless your evals say otherwise.

GPT-6 Astra (OpenAI) - September 2026 NEW

Attribute	Value
Model ID	`gpt-6-astra`
Context Window	1.05M tokens; 128K max output; knowledge cutoff April 30, 2026
Input / Output Cost	$10.00 /$ 50.00 per 1M; above 272K input the whole request bills at $20 /$ 75
Cache	Cached input $1.00; cache write$ 12.50; fixed 30-minute TTL
Service tiers	Batch and Flex 50% off; Fast 2x ( $20 /$ 100); Ultrafast 6x ( $60 /$ 300, GA September 29, Responses API)
Reasoning	Efforts `low`, `medium`, `high`, `xhigh`, `max`; no `none`; no custom `temperature`, `top_p`, or logprobs; tool calling requires the Responses API
Availability	Staged rollout from September 3 (Daybreak organizations first), in the API by September 5; Amazon Bedrock GA September 8 (In-Region and Geo inference $11 /$ 55, 10% over Global); Microsoft Foundry GA
Safety rating	First OpenAI model rated Critical for cybersecurity under its Preparedness Framework; High for biological and chemical capability. OpenAI says covert sandbagging is likely not reliably caught
Benchmarks	Terminal-Bench 4.0 leaderboard 58.18% at max (rank 1, tbench.ai, September 21). ARC-AGI-3 (ARC Prize): 62.7% on the provider-neutral Standard harness (max, $26,098) vs 99.9% on ARC Prize's Provider Adapter harness (high,$ 18,817), which uses OpenAI's own context-management features. AA Intelligence Index v4.3.2: 53 (max)
Released	September 3, 2026

What it is: OpenAI's new ceiling model, at the same $10/$ 50 as Claude Fable 5.1. Three things change cost models. First, speed is a priced axis with three rungs (Standard, Fast at 2x, Ultrafast at 6x), so a latency SLO is a budget decision. Second, the 272K cliff applies to the entire request, so a prompt just over the line costs roughly twice one just under it: prompt-length routing is a cost control. Third, token efficiency: Artificial Analysis measured Astra using about a third of GPT-5.6 Sol's tokens per coding task, so its cost per task sits closer to the mid tier than the list price suggests.

Read the harness: the 37-point ARC-AGI-3 swing on the same model and test set comes from whether the scaffold preserves opaque reasoning state between requests. Quote Astra's numbers with harness, effort, and who ran them.

GPT-6 Sol and GPT-6 Luna (OpenAI) - September 2026 NEW

Attribute	Value
Model IDs	`gpt-6-sol`, `gpt-6-luna` (text and image in, text out; Responses and Chat Completions)
Context Window	1.05M tokens; 128K max output
Sol pricing	$2.00 /$ 10.00 per 1M; cached input $0.20; cache write$ 2.50; above 272K $4 /$ 15
Luna pricing	$0.10 /$ 0.50 per 1M; cached input $0.01; cache write$ 0.125; above 272K $0.20 /$ 0.75
Batch / Flex	Sol $1 /$ 5; Luna $0.05 /$ 0.25
Lineage	Same types of data and training as GPT-6 Astra, per OpenAI. They replace GPT-5.6 Sol and Luna at about half the price; no GPT-6 Terra was announced
Availability	API and Amazon Bedrock GA September 22, 2026
Known issue	An image-encoding bug degraded image understanding and computer use until a September 25 fix shipped under the same model IDs; OpenAI told customers to rerun image evals
Released	September 22, 2026

Why it matters: Luna at $0.10/$ 0.50 is OpenAI's cheapest current-generation model. The retiring gpt-5-nano ( $0.05/$ 0.40) stays cheaper until it shuts down December 11, and OpenAI names gpt-5.6-luna ( $0.20/$ 1.20) as its replacement, so moving to Luna instead avoids a price rise. Luna undercuts DeepSeek V4.1-Flash even off-peak ( $0.15/$ 0.60). Gemini 2.5 Flash-Lite ( $0.10/$ 0.40, existing users only) and Meta's Muse Spark contributor tier ( $0.10/$ 0.20) are cheaper, and some hosted open models are cheaper on output (MiMo-V2.6-Flash $0.14/$ 0.28, Qwen3.8-Flash $0.15/$ 0.47). With no Terra successor, three-tier OpenAI routing designs collapse to two, or keep GPT-5.6 Terra at $2/$ 12, which GPT-6 Sol now undercuts on output. The image bug is the operational lesson: a provider-side fix changed behavior under a stable ID, so evals must rerun on vendor changelog events, not only on model-ID changes. OpenAI's Sol post reports GPT-6 Sol at xhigh 60.5% vs Claude Opus 5 at medium 60.3% on its offline OSWorld 2.0 set (vendor-reported).

GPT-6.1 Sol (OpenAI) - September 2026 NEW

Attribute	Value
Model ID	`gpt-6.1-sol`
Context Window	1.05M tokens; 128K max output; knowledge cutoff April 30, 2026
Input / Output Cost	$2.00 /$ 10.00 per 1M; above 272K $4 /$ 15
Cache	Cached input $0.10 (0.05x, half of GPT-6 Sol's rate); cache write$ 2.50
Reasoning	`low`, `medium` (default), `high`, `xhigh`, `max`
Features	Hosted shell, computer use, MCP, multi-agent delegation (beta, Responses API)
Availability	API; Amazon Bedrock, where it follows OpenAI's lifecycle and deprecation terms rather than a Bedrock minimum-availability date
Benchmarks	AA Intelligence Index v4.3.2: 52 (max)
Released	September 29, 2026

What it is: OpenAI's model for complex coding and professional work at a lower cost than Astra, shipped one week after GPT-6 Sol at the same list price. Codex CLI switched its default to GPT-6.1 Sol in rust-v0.159.1 the same day, a silent default swap for anyone who had not pinned a model. Note the Bedrock terms: cross-cloud availability no longer implies a cloud-guaranteed support window.

GPT-5.4 (OpenAI) - March 2026

Attribute	Value
Context Window	272K tokens (standard); extended available
Input Cost	$2.50 / 1M tokens (cached$ 0.25)
Output Cost	$15.00 / 1M tokens
Multimodal	Text, Vision, native computer use
Highlights	Built-in computer-use capabilities; 33% fewer factual errors vs GPT-5.2; combines coding + agentic strengths
Released	March 2026

Best for: Existing agentic workflows already validated on it. Considerations: Above 272K input tokens the whole request bills at long-context rates ( $5 /$ 22.50). GPT-6 Sol is cheaper ( $2/$ 10) for new builds.

GPT-5.4-mini (OpenAI) - March 2026

Attribute	Value
Context Window	272K tokens
Input Cost	$0.75 / 1M tokens
Output Cost	$4.50 / 1M tokens
Highlights	Cost/performance option for high-volume GPT-5 tier workloads
Released	March 2026

Best for: High-volume API calls already tuned on it. For new volume tiers, GPT-6 Luna ( $0.10/$ 0.50) is far cheaper. GPT-5.4-nano is $0.20 /$ 1.25 and retires April 1, 2027.

GPT-5.4 Pro (OpenAI) - March 2026

Attribute	Value
Context Window	272K tokens
Input Cost	$30.00 / 1M tokens
Output Cost	$180.00 / 1M tokens
Highlights	Maximum reasoning power; premium tier for hardest tasks
Released	March 2026

Best for: Competition-level math, complex multi-step reasoning. Considerations: Very expensive. GPT-5.4 Pro is still listed at $30 /$ 180, the same as GPT-5.5 Pro, the newer dedicated Pro ID. On the GPT-5.6 and GPT-6 families Pro is a mode (reasoning.mode: "pro") billed at the selected model's standard rates.

GPT-5.6 Sol / Terra / Luna (OpenAI) - GA July 9, 2026 (Sol and Luna superseded by GPT-6)

Attribute	Value
Variants	Sol (flagship), Terra (balanced), Luna (fast, low cost)
Context Window	1.05M tokens (all three); 128K max output; knowledge cutoff February 16, 2026
Sol pricing	$5.00 /$ 30.00 per 1M list; promotional $4 /$ 20 since August 21, 2026, guaranteed only "at least through November 21, 2026", so budget on list
Terra pricing	$2.00 /$ 12.00 per 1M (cut 20% from $2.50/$ 15 on July 30)
Luna pricing	$0.20 /$ 1.20 per 1M (cut 80% from $1/$ 6 on July 30)
Long context	Above 272K input the whole request bills at Sol $8 /$ 30 (promotional), Terra $4 /$ 18, Luna $0.40 /$ 1.80
Speed tiers	Fast (renamed from Priority on July 30) at 2x; Ultrafast at 6x in preview on Sol
Reasoning	"max" reasoning effort plus an "ultra" mode that uses subagents to accelerate complex work
API features at GA	Programmatic tool calling, multi-agent support, explicit prompt-cache breakpoints
Benchmarks	Sol: 53.6 on Agents' Last Exam (13.1 points ahead of Claude Fable 5) and a Terminal-Bench 2.1 record. Claude Fable 5 led SWE-Bench Pro (80 vs Sol's 64.6), a benchmark OpenAI publicly disputes. Vendor-reported. On the Terminal-Bench 4.0 leaderboard, Sol scores 37.27% at max
Released	Limited preview June 26, 2026; general availability July 9, 2026

What it is: OpenAI's first flagship line shipped as three models. The June 26 preview was gated at the request of the US government over dual-use cybersecurity capability; GA followed a 13-day review. The three-tier structure plus the July 30 cuts reset routing math for tiered model selection.

Status (October 2026): GPT-6 Sol and Luna replace Sol and Luna at about half the price. Terra has no GPT-6 successor and stays at $2/$ 12. This family will stay in production for a while anyway: Foundry moves its retiring o-series deployments to GPT-5.6 Sol and Terra on November 19, and OpenAI's December 11 retirement of the dated GPT-5 and o3 snapshots points to the GPT-5.6 models.

GPT-5.6-Cyber (OpenAI) - August 2026 (restricted)

Attribute	Value
Model ID	`gpt-5.6-cyber`
Context Window	400K total (272K max input, 128K max output)
Input / Output Cost	$12.50 /$ 75.00 per 1M; cached input $1.25
Access	Daybreak Red tier only, with identity verification, legal attestations, and approved use cases. Responses API only. Hardware security keys were scheduled to become mandatory on individual Daybreak accounts from September 1, 2026 (enforcement not independently confirmed)
Refusal posture	Trained for a lower refusal rate on dual-use security work: 95.0% completion on OpenAI's internal Advanced Cybersecurity Completion Rate eval versus 1.5% for GPT-5.6 Sol
Released	August 10, 2026 (the Daybreak Blue and Red tier split appears in the API changelog dated August 7)

What it is: A cybersecurity model built on GPT-5.6 Sol, shipped alongside a split of the Daybreak program into two tiers. Daybreak Blue gives approved defenders access to general-purpose frontier models for vulnerability discovery, secure code review, detection engineering, and incident response. Daybreak Red is separately approved and gates gpt-5.6-cyber for vulnerability reproduction, exploit validation, penetration testing, and red teaming.

Why it matters architecturally: A model that is deliberately more permissive than the frontier default, priced at 2.5x Sol's list price, restricted to a single API surface, and fenced behind identity verification is a reference design for how labs operationalize dual-use access. Gated models also retire fast: OpenAI deprecated the earlier gpt-5.4-cyber on September 11 and shut it down October 1, about 20 days' notice against its stated 3-month minimum for specialized variants.

Capability-Tiered Access Programs (October 2026)

Lab	Program	Gated model(s)	Who gets in
Anthropic	Project Glasswing; Life Sciences Verification Program (first participants enrolled); Cyber Verification Program (exists today for Opus- and Sonnet-class models with reduced cyber safeguards; Mythos-class access announced for "the near future")	Claude Mythos 5.1 (same weights as Fable 5.1, looser safeguards)	Glasswing partners; life-sciences participants enrolled with the US government; CVP members once Mythos-class access is added (Mythos 5.1 is currently limited to a set of US organizations)
OpenAI	Daybreak Blue / Daybreak Red	Blue: general frontier models (Daybreak organizations got GPT-6 Astra first); Red: `gpt-5.6-cyber`	Approved defenders; Red is separately approved
OpenAI	Trusted access	`gpt-rosalind-research` (life sciences; GA September 8 at $5 /$ 25 per 1M, billed from October 5)	Approved life-sciences organizations
Google	Fairwind Program	Gemini 3.8 Flash Cyber; first access to Gemini 4 Argon, released to them without cyber guardrails	Government authorities, critical infrastructure operators, software maintainers

Design implication: capability-tiered access is now an industry pattern, not a one-off. Treat a gated tier as a separate model with its own identity requirements, API surface, retention terms, and a much shorter retirement runway.

GPT-5.5 (OpenAI) - April 2026

Attribute	Value
Context Window	1M tokens
Input Cost	$5.00 / 1M tokens (cached$ 0.50)
Output Cost	$30.00 / 1M tokens
Long context / Fast	Above 272K input: $10 /$ 45. Fast is 2.5x ( $12.50 /$ 75), not 2x as on newer models
Multimodal	Text, Image, Audio, Video
ARC-AGI-2	85.0% (leader as of May 13, 2026)
Released	April 23, 2026

Best for: Multimodal workloads already validated on it. Pitched at launch as a "new class of intelligence for real work", it replaced GPT-5.4 for top-tier reasoning and multimodal work. Considerations: ~2× the input cost of GPT-5.4 ( $2.50 →$ 5.00) and ~2× output ( $15 →$ 30). For new builds, GPT-6 Sol ( $2/$ 10) is cheaper and newer.

GPT-5.5 Instant (OpenAI) - May 2026

Attribute	Value
Status	Became the ChatGPT default and `chat-latest` in the API on May 5, 2026
Hallucination Reduction	52.5% fewer on high-stakes prompts (medicine/law/finance) vs GPT-5.3 Instant
AIME 2025	81.2% (up from 65.4% on GPT-5.3 Instant)
Response Length	~30% fewer words/lines than predecessor
Released	May 5, 2026

Best for: ChatGPT-equivalent chat workloads, high-stakes domains where hallucination reduction matters. Considerations: Replaced GPT-5.3 Instant as the chat default. GPT-5.2-chat-latest and GPT-5.3-chat-latest deprecated May 8, 2026.

OpenAI Voice Models: GPT-Live-1, GPT-Realtime-2.1, Transcription (October 2026)

Model	Shape	Pricing	Status
`gpt-live-1`	Full-duplex voice front end that delegates reasoning and tools to a backend (an OpenAI Responses model or your own) on `v1/live/sessions`	$0.05 per session minute, billed per second, plus backend model tokens	GA September 10, 2026
`gpt-realtime-2.1` / `gpt-realtime-2.1-mini`	Single speech-to-speech model on the Realtime API; turn-based with `server_vad` or `semantic_vad`; configurable reasoning effort	2.1: audio $32 in /$ 0.40 cached / $64 out per 1M; mini:$ 10 / $0.30 /$ 20	July 6, 2026 (succeeds `gpt-realtime-2`, May 7)
`gpt-realtime-translate`	Streaming translation, one session per target language	$0.034 per minute	Current
`gpt-transcribe` / `gpt-live-transcribe`	Batch / streaming speech-to-text	$0.0045 /$ 0.017 per minute	July 28, 2026
`gpt-realtime`, `gpt-realtime-mini`, `gpt-4o-realtime`, `gpt-audio`	Legacy speech-to-speech and audio	-	Shut down January 20, 2027 (to `gpt-realtime-2.1`, `-2.1-mini`, or `gpt-audio-1.5`)
`whisper-1`, `gpt-4o-transcribe` family	Legacy speech-to-text	-	Shut down February 26, 2027 (to `gpt-transcribe` or `gpt-live-transcribe`)
`tts-1`, `tts-1-hd`, `gpt-4o-mini-tts` snapshots	Legacy text-to-speech	-	Shut down January 6, 2027 (announced October 1; to `gpt-realtime-2.1-mini`)

Three architectures, not two. OpenAI's docs now describe full-duplex-plus-delegation (GPT-Live), a single speech-to-speech model (Realtime API), and a chained STT, LLM, TTS pipeline. GPT-Live separates the voice layer from the reasoning layer, so business rules and authoritative task state belong in the backend. Once its 128K context passes 90% full, it restarts the voice engine with the instructions plus at most 8,192 tokens of history (including a summary of older turns), so keep slots and bookings in application state. Realtime sessions cap at 60 minutes and re-send the conversation on every response, so cost grows with call length. OpenAI reports GPT-Live turn-taking latency of 0.798 s vs 1.41 s for gpt-realtime-2.1 (vendor-reported, and dependent on the backend you pick). The Realtime API beta interface was removed May 12, 2026.

Other vendors: Google's gemini-3.8-live and gemini-3.8-live-extended-thinking went GA September 15 (audio $3 in /$ 12 out per 1M; asynchronous tool calls by default, and the Extended Thinking variant reasons in the background while it talks). Google reports tau-Voice 68.6% for Extended Thinking against 30.1% for the base model (vendor-reported), which is the pattern to design for: voice task success tracks the reasoning attached to the voice layer, and that reasoning has to fit the turn latency budget. gemini-3.8-flash-tts went GA September 22 at a promotional $0.50 /$ 9 per 1M through December 31 ( $1 /$ 18 from January 1, 2027). xAI's grok-voice-think-fast-2.0 speaks OpenAI-Realtime-compatible events at $0.08 per minute. Anthropic has no speech API: Claude can only be the text backend in a voice stack, for example behind GPT-Live client delegation. See Realtime Voice Agents.

Gemini 4 Argon (Google) - ANNOUNCED September 30, 2026, NOT GA

Attribute	Value
Status	Announced September 30, 2026. Rolling out first to trusted cyber defenders in the Fairwind Program (without cyber guardrails); paid API customers and Google AI Ultra subscribers come next, with no date given
Model ID	None published; not on the Gemini API models page or changelog as of October 1
Announced pricing	Introductory $2 /$ 10 per 1M (no end date given), then $4 /$ 20; cached input 95% off
Output limit	1M tokens, up from 64K (Google)
Benchmarks	Google-reported: DeepSWE v1.1 77.9%, AutomationBench 51.3%, LVBench 91.7%. AA Intelligence Index v4.3.2: 53 (high, scored before general availability)
Target workloads	Real-world software engineering, enterprise knowledge work (legal, finance), cyber defense

How to treat it: as a planning input, not a dependency. Argon is a new Gemini 4 line, not the long-delayed Gemini 3.5 Pro. If the 1M-token output limit reaches the API, whole-repo rewrites and long document generation stop needing chunked generation loops, which is a real architecture change. Until there is a model ID, gemini-3.1-pro-preview remains Google's only Pro-tier API model.

Gemini 3.8 Flash (Google) - September 2026 NEW

Attribute	Value
Model ID	`gemini-3.8-flash` (stable, GA)
Context Window	1,048,576 tokens in / 65,536 out
Input / Output Cost	$0.75 /$ 3.75 per 1M through December 31, 2026; $1.50 /$ 7.50 from January 1, 2027. Batch and Flex 50% off; Priority $1.35 /$ 6.75 ( $2.70 /$ 13.50 from January 1); cache read $0.075 per 1M plus storage
Modalities	Text, image, video, audio, PDF in; text out
Thinking	`thinking_level` `low`, `medium` (default), `high`; no minimal level; replaces `thinking_budget`
Tools	Function calling, structured outputs, caching, code execution, computer use (preview), file search, Search and Maps grounding; no Live API
Benchmarks	Google-reported: DeepSWE v1.1 73.7% vs Claude Opus 5 74.0%. Independent: AA Intelligence Index v4.3.2 41 (high); SWE-Bench Pro v2 private set 211/272 (Scale)
Released	September 2, 2026

What it is: Further training on Gemini 3.7 Flash rather than a new base, and Google's third Flash in six weeks. It is the Gemini model Google recommends for computer use. The per-token price is low, but coverage notes it spends more reasoning tokens per task. On DeepSWE v1.1, where it tied Astra and Opus 5 near 74%, its cost per task was $2.36 at high effort versus$ 4.43 for Astra at xhigh and $11.84 for Opus 5 at max. Model the January 2027 price before committing volume: 3.6, 3.7, and 3.8 Flash all double on the same day.

Gemini 3.8 Flash Cyber: announced the same day for trusted defenders in the new Fairwind Program, with more permissive cybersecurity mitigations and Google's CodeMender harness for find-verify-fix pipelines (Google reports CyberGym 86.2%). It replaces Gemini 3.5 Flash Cyber in Google's tiered-access lineup.

Gemini 3.1 Pro Preview (Google) - February 2026

Attribute	Value
Model ID	`gemini-3.1-pro-preview` (Preview)
Context Window	1M tokens
Input Cost	$2.00 / 1M tokens (standard);$ 4.00 (200K+)
Output Cost	$12.00 / 1M tokens (standard);$ 18.00 (200K+)
Multimodal	Native: Text, Vision, Audio, Video
Highlights	Google's top Pro-tier API model; strong agentic and coding capabilities
Released	February 2026

Best for: Complex reasoning, multimodal analysis, long-context workloads. Considerations: Replaced Gemini 3 Pro Preview. As of October 1 it is the only 3-series Pro model on the Gemini API; there is no Gemini 3.5 Pro. Gemini 2.5 was not deprecated in June as once scheduled: on September 18 Google restricted the 2.5 models to users who had already used them, said they "are not deprecated", and pointed new projects to 3.5 Flash-Lite or 3.8 Flash. Google Cloud (Vertex AI) retires gemini-2.5-pro, gemini-2.5-flash, and gemini-2.5-flash-lite on October 20, 2026.

Gemini Flash and Flash-Lite Lineup (Google)

Model	Input / Output per 1M	Status
`gemini-3.8-flash`	$0.75 /$ 3.75 intro; $1.50 /$ 7.50 from January 1, 2027	GA September 2, 2026; the current default
`gemini-3.7-flash`, `gemini-3.6-flash`	Same intro pricing as 3.8 Flash	Superseded by 3.8 Flash
Gemini 3.5 Flash	$1.50 /$ 9.00	Available
Gemini 3.5 Flash-Lite	$0.30 /$ 2.50	Google's pointer, with 3.8 Flash, for projects leaving 2.5
Gemini 3.1 Flash-Lite	$0.25 /$ 1.50	Shuts down May 7, 2027 (to 3.5 Flash-Lite)
Gemini 3 Flash Preview	$0.50 /$ 3.00	Preview
Gemini 2.5 Flash-Lite	$0.10 /$ 0.40	Existing users only since September 18

Watch for a phantom row: a "Gemini 3.1 Flash" at $0.10 /$ 3.00 still circulates in older cost tables. No model with that name and price appears on Google's pricing page; the nearest are Gemini 3 Flash Preview and Gemini 3.1 Flash-Lite. Price new work on Gemini 3.8 Flash.

Gemini 3.2 Flash (Google) - May 2026, Historical

Attribute	Value
Status	Quiet rollout in iOS Gemini app and Google AI Studio May 5, 2026 (no formal announcement)
Released	May 5, 2026

Status: Historical. Google's Flash line has since moved through 3.5, 3.6, 3.7, and 3.8; build on gemini-3.8-flash.

Gemini Deep Research / Deep Research Max (Google) - April 2026

Attribute	Value
Built on	Gemini 3.1 Pro
Capabilities	MCP support; native chart/infographic generation; extended test-time compute; async background workflows
Released	April 21, 2026

Best for: Research agents, document synthesis, long-running async workflows. The MCP support made it the first Google research-agent product with first-class tool integration.

Gemini Robotics-ER 1.6 (Google DeepMind) - April 2026

Attribute	Value
Domain	Physical robotics, embodied reasoning
New capability	Reading gauges/sight glasses
Deployment	Boston Dynamics Spot
Released	April 14, 2026

Best for: Robotics applications requiring vision-language grounding for physical actions. Available via Gemini API and AI Studio.

Gemini 3.7 Flash (Google) - August 2026, SUPERSEDED by 3.8 Flash

Attribute	Value
Model ID	`gemini-3.7-flash`
Context Window	1,048,576 tokens in / 65,536 out
Input / Output Cost	$0.75 /$ 3.75 per 1M through December 31, 2026, then $1.50 /$ 7.50. Context caching $0.075 per 1M; Batch$ 0.375 / $1.875
Multimodal	Text, image, audio, video; agentic video understanding is opt-in (`"processing": "agentic"`), not on by default
Knowledge cutoff	March 2026
Released	August 13, 2026 (GA)

What it is: Google's workhorse tier, built on Gemini 3.6 Flash with what the model card describes as algorithmic improvements to the reasoning foundation rather than a new pretraining run. Customizable thinking configurations trade quality against cost and latency per request. Google-reported gains over 3.6 Flash include FrontierCode 1.1 Main 43.6% versus 34.4% and WebDev Arena Elo 1588 versus 1538.

Agentic video (September 1): with "processing": "agentic" on a video input, the model decides which segments to inspect, at what speed, and through which modality, instead of sampling at a fixed 1 FPS (still the default). Google reports up to 88% fewer tokens and up to 66% lower cost at standard token pricing, and the mode now covers 3.8 Flash, 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. Video cost models built on fixed-FPS sampling can overestimate tokens by as much as roughly 8x.

Considerations: Superseded by 3.8 Flash on September 2 at the same intro price. Gemini 3.5 Pro still had not shipped by October 1; Google's next flagship was announced as Gemini 4 Argon instead.

Grok 4.7 (SpaceXAI) - September 2026 NEW

Attribute	Value
Model ID	`grok-4.7`
Context Window	500K tokens; text and image in, text out
Input / Output Cost	$2.00 /$ 6.00 per 1M below a 200K prompt; $4.00 /$ 12.00 on all tokens once the prompt reaches 200K. Cached input $0.50 ($ 1.00 at 200K+)
Fast variant	About 2x output speed at 2x price, offered only in Cursor and Grok Build (not a listed API model)
Knowledge cutoff	May 2026 per the API docs
Benchmarks	Terminal-Bench 4.0 leaderboard 37.58% at xhigh. AA Intelligence Index v4.3.2: 46, up from 44 for Grok 4.6
Released	September 21, 2026 (Grok API, Grok Build, Cursor on all plans, OpenRouter and other gateways)

What it is: Unlike most August releases, a new and larger base model, with a longer RL run weighted toward multi-hour tasks (per SpaceXAI). It is also this quarter's clearest case of vendor tables outrunning independent indices: large gains in SpaceXAI's own table, about 2 points on AA. Distribution now runs through Cursor, which became part of SpaceX in mid-August, and Grok launches appear on Cursor's blog. The whole-request 200K pricing trap carries over from 4.6.

Grok 4.6 (SpaceXAI) - August 2026, SUPERSEDED by Grok 4.7

Attribute	Value
Model ID	`grok-4.6`
Context Window	500K tokens
Input / Output Cost	$2.00 /$ 6.00 per 1M below a 200K prompt; $4.00 /$ 12.00 at or above 200K. Cached input $0.50
Reasoning	Effort settings low, medium, high (default), xhigh
Knowledge cutoff	February 1, 2026
Released	August 12, 2026

What it is: SpaceXAI's frontier model before 4.7, focused on long-running agents and interactive visual work. It scored 61 on the Artificial Analysis Intelligence Index at launch, up from 56 for Grok 4.5 High, but that was on the pre-September index; on the re-baselined v4.3.2 it scores 44, and the two scales cannot be compared.

Two traps worth knowing. The higher long-prompt rate applies to all tokens in the request once the prompt reaches 200K, not just the tokens past the threshold. And cached input got more expensive than Grok 4.5 ( $0.50 versus$ 0.30 per 1M), so cache-heavy agent loops do not automatically get cheaper on the upgrade. Note the vendor name: xAI completed a rebrand to SpaceXAI, so current docs and release notes use the new name.

Grok 4 (xAI) - RETIRED on the xAI API

Attribute	Value
Context Window	256K tokens
Input Cost	$3.00 / 1M tokens
Output Cost	$15.00 / 1M tokens
Highlights	Native tool use and real-time search; competitive reasoning
Released	July 2025 (Grok 4.20 beta: February 2026)

Status: Retired from the xAI API on May 15, 2026, along with grok-4-fast, Grok 4.1 Fast, grok-code-fast-1, and grok-3. The retired text-model slugs still resolve but redirect to grok-4.3 ( $1.25 /$ 2.50) and bill at its rates (grok-code-fast-1 goes to grok-build-0.1), so code keeps "working" while behavior and cost change. Azure Foundry still lists grok-4 and grok-4-1-fast as GA, a reminder that lifecycle is tracked per (model, platform) pair.

MAI-Thinking-1 (Microsoft) - August 2026 NEW (public preview)

Attribute	Value
Status	Public preview on Microsoft Foundry since August 12, 2026
Architecture	Sparse MoE, about 1T total / 35B active parameters; 256K context
API	Chat Completions compatible, function calling
Pricing	Not published
Benchmarks	Microsoft-reported: AIME 2025 97.0%, SWE-Bench Pro on par with Claude Opus 4.6, blind human preference over Claude Sonnet 4.6 across 1,276 tasks

Why it matters: Microsoft now has a first-party reasoning model in Foundry next to its OpenAI and Anthropic offerings, which matters for vendor diversification and sovereignty discussions. Microsoft says it did not distill from other labs' models. Do not cost a design on it until a price is published.

Model Comparison: Frontier and Workhorse Tiers (October 2026)

Model	Input / Output per 1M	Cache read	Context	Access	Status
Claude Fable 5.1	$10 /$ 50	$0.25 (0.025x)	1M, flat pricing	5 platforms; ZDR only if authorized	GA September 1
GPT-6 Astra	$10 /$ 50 ( $20 /$ 75 above 272K)	$1.00	1.05M	API, Bedrock, Foundry	Staged GA from September 3
Claude Opus 5.5	$4 /$ 20	$0.20 (0.05x)	1M, flat pricing	5 platforms; ZDR	GA September 22
Gemini 4 Argon	$2 /$ 10 intro, then $4 /$ 20	95% off	1M output limit (announced)	Fairwind only	Announced September 30, not GA
Claude Sonnet 5.5	$2 /$ 10	$0.20	1M, flat pricing	5 platforms; ZDR	GA September 28
GPT-6.1 Sol	$2 /$ 10 ( $4 /$ 15 above 272K)	$0.10 (0.05x)	1.05M	API, Bedrock	GA September 29
GPT-6 Sol	$2 /$ 10 ( $4 /$ 15 above 272K)	$0.20	1.05M	API, Bedrock	GA September 22
GPT-5.6 Terra	$2 /$ 12 ( $4 /$ 18 above 272K)	$0.20	1.05M	API, Foundry	GA July 9; no GPT-6 successor
Grok 4.7	$2 /$ 6 ( $4 /$ 12 at 200K+)	$0.50	500K	xAI API, Cursor, gateways	GA September 21
Muse Spark 1.3	$1.25 /$ 4.25	$0.15	1M	Meta Model API, Muse Code	GA September 2
Gemini 3.8 Flash	$0.75 /$ 3.75 intro ( $1.50 /$ 7.50 from January 1)	$0.075	1M in / 65,536 out	Gemini API, Gemini Enterprise	GA September 2
GPT-6 Luna	$0.10 /$ 0.50 ( $0.20 /$ 0.75 above 272K)	$0.01	1.05M	API, Bedrock	GA September 22

"5 platforms" = Claude API, Bedrock, Google Cloud, Microsoft Foundry, Claude Platform on AWS. Standard tier, list prices. Benchmarks for these models are in Capability Comparison; full pricing mechanics (batch, speed tiers, residency premiums) are in Pricing and Costs.

Production Heritage & Maturity

While frontier models lead on benchmarks, many enterprise systems rely on battle-tested models, and most of them now have a retirement date:

Model Family	Production Since	Maturity Note
GPT-4o	May 2024	Most mature ecosystem. The original `gpt-4o-2024-05-13` snapshot shuts down on the OpenAI API on October 23, 2026 (to GPT-5.6 Sol). On Microsoft Foundry, Standard deployments of it auto-upgrade to GPT-5.6 Sol on December 9: a reasoning-class model swapped in with no code change.
Claude Sonnet 4.5 / 4.6	Sept 2025 / Feb 2026	Inherited the tool-use reliability reputation of Claude 3.5 and 3.7 Sonnet. Sonnet 4.5 retires November 30, 2026 (Claude API and Foundry); Sonnet 4.6 is legacy and now costs more than Sonnet 5.5.
Gemini 2.5 Pro	March 2025	Proven long-context. Restricted to existing users on the Gemini API since September 18, 2026 ("not deprecated", no shutdown date); Google Cloud (Vertex) retires it October 20, 2026.
o1 / o3	Sept 2024	Well-understood reasoning failure modes. The OpenAI API shuts down `o1`, `o1-pro`, `o3-mini`, and `o4-mini` on October 23; Foundry retires o1, o3, o3-pro, and siblings on November 19 (to GPT-5.6 Sol); OpenAI shuts down `o3-2025-04-16` and the dated GPT-5 snapshots on December 11.

Why stay on "older" frontier models?

Consistency: New models have release-window latency spikes and behavior shifts. GPT-6 Sol's image path was degraded until a fix three days after launch.
Cost is no longer the reason: Below the ceiling tier (where GPT-6 Astra costs more than GPT-5.6 Sol), September's new generation was cheaper. Opus 5.5 undercuts Opus 5 by 20%, GPT-6 Sol and Luna cost roughly half of GPT-5.6 Sol and Luna, and Sonnet 5.5 costs less than Sonnet 4.6. Staying put now costs money.
Guardrail Tuning: Security and moderation layers are more refined.
The runway is short: OpenAI gives 6 months' notice for GA models, 3 months for specialized variants, and as little as 2 weeks for previews; Anthropic gives at least 60 days; new Bedrock models get 6 months or 45 days. Pin dated IDs and track retirements per (model, platform) pair.

Retirement Calendar (as of October 1, 2026)

Date	What retires	Move to
October 5	`antigravity-preview-05-2026`	`antigravity-preview-09-2026`
October 8	Bedrock Claude Opus 4.1 enters higher-priced public extended access	Opus 5.5
October 14	Bedrock Claude Sonnet 4; Foundry `gpt-4.1-nano`	Sonnet 5.5; a current small OpenAI model
October 15	Claude Haiku 4.5 retirement floor on the Claude API (no notice yet, so not before about December); Foundry `gpt-4o-transcribe` and `gpt-4o-mini-transcribe` (2025-03-20)	Haiku 5.5 once released; `gpt-transcribe` or `gpt-live-transcribe`
October 20	Google Cloud (Vertex) `gemini-2.5-pro`, `-flash`, `-flash-lite`	Gemini 3.8 Flash or 3.5 Flash-Lite
October 23	OpenAI API legacy snapshots (announced April 22): `gpt-4o-2024-05-13`, `gpt-4-turbo`, `gpt-4-0613`, `gpt-3.5-turbo-0125`, `gpt-4.1-nano`, `o1`, `o1-pro`, `o3-mini`, `o4-mini`, `gpt-image-1`, `gpt-4-1106-preview`, the bare `gpt-4` and `gpt-3.5-turbo` aliases, and fine-tunes of GPT-3.5 Turbo, GPT-4, GPT-4.1-nano, o4-mini, babbage-002 and davinci-002	GPT-5.6 Sol, Terra, or Luna; `gpt-image-2.5-sunburst` or `-flare` for images
October 31	Mistral La Plateforme GLM 5.2; OpenAI Evals become read-only	GLM 5.3; export eval results before shutdown
November 2	`grok-imagine-image-quality`	`grok-imagine-image-2.0`
November 15	Foundry `claude-haiku-4-5` and `codex-mini`	-
November 19	Foundry o1, o1-pro, o3, o3-pro, o3-deep-research; o3-mini, o4-mini	GPT-5.6 Sol; GPT-5.6 Terra
November 24	Foundry `claude-opus-4-5`	-
November 26	Bedrock AI21 Jamba 1.5	-
November 30	`claude-sonnet-4-5-20250929` (Claude API, Foundry); OpenAI Evals, Agent Builder, and Reusable Prompts	`claude-sonnet-5-5`; Promptfoo (OpenAI-owned, MIT) for evals
December 1	`gpt-image-1.5`, `gpt-image-1-mini`, `chatgpt-image-latest`	`gpt-image-2.5-sunburst` or `gpt-image-2.5-flare`
December 9	Foundry gpt-4o 2024-05-13 auto-upgrade	GPT-5.6 Sol
December 11	`gpt-5-2025-08-07`, `o3-2025-04-16`, and sibling snapshots	GPT-5.6 Sol, Terra, or Luna
December 15	Foundry `whisper`, `tts`, `tts-hd`	Current OpenAI speech models
January 6, 2027	`tts-1`, `tts-1-hd`, `gpt-4o-mini-tts` snapshots; OpenAI fine-tuning job creation ends	`gpt-realtime-2.1-mini` for speech
January 8, 2027	Bedrock Claude Opus 4.1	Opus 5.5
January 20, 2027	`gpt-realtime`, `gpt-audio`, and the gpt-4o realtime/audio families	`gpt-realtime-2.1`, `gpt-audio-1.5`
February 26, 2027	`whisper-1`, `gpt-4o-transcribe` family	`gpt-transcribe`, `gpt-live-transcribe`
April 1, 2027	`gpt-5.1`, `gpt-5.3-codex`, `gpt-5.4-nano`	GPT-6 Sol or GPT-6 Luna

Already past (August 15 to October 2):

Executed: Groq free and developer tiers' Llama 3.3 70B and Llama 3.1 8B (August 16), Anthropic Workbench (August 17), OpenAI and Azure Assistants API (August 26, to Responses plus Conversations), Mistral Medium 3.1 (August 31, to Medium 3.5), OpenAI Sora 2 and the Videos API (September 24, no replacement), and gpt-5.4-cyber (October 1, on 20 days' notice).
Scheduled on vendor lifecycle pages, execution not independently confirmed: Bedrock Claude 3 Haiku (September 10); Bedrock Amazon Nova Premier and Nova Sonic v1 (September 14) and Nova Canvas and Reel (September 30); gpt-3.5-turbo-instruct, babbage-002, and davinci-002 (September 28); Mistral OCR 4.0 and Leanstral 1.5 on the API (September 30); and gemini-2.5-flash-image (October 2).

Open Source Models

Open-Weight Standings (October 2026)

Artificial Analysis Intelligence Index v4.3.2. The index was re-baselined on September 7, so do not compare these with pre-September figures (Kimi K3's 57.1 at launch was on an older version).

Model	AA v4.3.2	Total / active	License
Xiaomi MiMo-V2.6-Pro	46	1.02T / 42B	MIT
Z.ai GLM-5.3	45	744B / 40B	GLM-5.3 License (MIT plus a MaaS security review above US$10B revenue)
Moonshot Kimi K3	44	2.8T / 104B	Kimi K3 License (MaaS agreement above US$20M revenue)
Z.ai GLM-5.3-Flash	42	320B / 18B	MIT
Alibaba Qwen3.8 2.4T-A95B	40 (at v4.3 launch)	2.4T / 95B	Qwen3.8-Max License
Alibaba Qwen3.8-Flash-Next	40	125B / 6B (plus 51B N-gram memory)	Qwen Community License 1.0
DeepSeek V4.1-Flash	39	552B backbone; 8B prefill / 16B decode active	MIT
Alibaba Qwen3.8-27B	34	27B dense	Apache 2.0
MBZUAI K2 Horizon 375B-A23B	31	375B / 23B	Apache 2.0, open training data
MiniMax-M3	29	~428B / 23B	minimax-community
Thinking Machines Inkling-Small	26	~266B / 12B	Apache 2.0
Thinking Machines Inkling	25	975B / 41B	Apache 2.0
NVIDIA Nemotron 3 Ultra	23	550B / 55B	OpenMDW-1.1
Meta Muse Glimmer	17	30B dense	Apache 2.0
Mistral Medium 3.5	14	128B dense	Modified MIT (no rights above US$20M monthly revenue)

For scale, the top closed model on the same index is Claude Opus 5.5 at 58. StepFun's Step 5 Preview also scores 44 but is API-only, so AA lists it as proprietary. Inkling-Small is the top US open-weight model. Outside China, South Korea's Motif 3 (34) leads, ahead of the UAE's K2 Horizon 375B (31).

Architecture note: the new open flagships all ship non-standard attention or KV schemes: sliding-window plus global layers on MiMo-V2.6-Pro, linear plus sparse attention on GLM-5.3-Flash and Qwen3.8-Flash-Next, sparse attention on Tencent Hy4, and a causal encoder-decoder with 890 bytes of global KV per token on DeepSeek V4.1-Flash. The textbook KV-size formula no longer transfers between models, so size capacity from each model card (see KV Cache and Context Caching). Speculative-decoding heads now ship inside checkpoints too: DSpark in DeepSeek V4-Pro-0813, MTP heads in MiMo-V2.6 and Qwen3.8-Flash-Next.

Llama 4 Family (Meta) - April 2025

Model	Parameters	Context	Architecture	Notes
Llama 4 Scout	17B active / 16 experts, 109B total (MoE)	10M	Sparse MoE	Industry-leading 10M context; fits a single H100 at Int4; beat Gemma 3 and Gemini 2.0 Flash-Lite per Meta
Llama 4 Maverick	17B active / 128 experts, 400B total (MoE)	1M	Sparse MoE	Beat GPT-4o and Gemini 2.0 Flash per Meta; comparable to DeepSeek V3 at half the active params
Llama 4 Behemoth	~288B active, about 2T total	-	Sparse MoE	Never released; Meta paused it in 2026, and its frontier models now ship closed as Muse Spark

Strengths:

First Llama generation with Mixture-of-Experts architecture
Natively multimodal from the ground up (text, image, video input)
Open weights on Hugging Face under the Llama 4 Community License; available via Meta AI on WhatsApp, Messenger, Instagram
Scout's 10M token context window is industry-leading for open models

Watch out: there is no Llama 4 8B, 70B, or 405B. Dense worked examples should use Llama 3.3 70B or Llama 3.1 405B, or a current open model. Fast-inference hosts are also dropping Llama: Groq retired Llama 4 Scout (July 17) and Llama 3.3 70B and Llama 3.1 8B (August 16) on its free and developer tiers.

Llama 3.x Family (Meta) - Previous Generation

Model	Parameters	Context	License	Notes
Llama 3.3 70B	70B	128K	Llama 3.3	Still widely deployed; strong general model
Llama 3.1 405B	405B	128K	Llama 3.1	Largest dense Meta model; superseded by Llama 4

Note: Llama 3.x remains in production, but Llama 4 Scout/Maverick offer better performance with lower active parameter counts thanks to MoE.

DeepSeek Family

Model	Parameters	Context	Status	Notes
DeepSeek V4.1-Flash	552B backbone (763B checkpoint); 8B active for prefill, 16B for decode (MoE)	1M (384K max output)	GA, September 10, 2026	MIT; API name `deepseek-flash`. New causal encoder-decoder: the decoder's global KV is projected from encoder states, plus Compressed Sparse Attention 2 and FP4 KV, for 890 bytes of global KV per token (about 1/4 of V4-Flash, per DeepSeek). Native image input; reasoning effort 1-100. API: peak $0.30 /$ 1.20 per 1M, cache hit $0.006; off-peak half. AA v4.3.2: 39.
DeepSeek V4 Pro	1.6T total / 49B active (MoE)	1M	GA (build V4-Pro-0813)	Previewed April 24, 2026. Uses ~27% compute / 10% memory of V3.2 at 1M tokens. SWE-bench Verified 80.6%. NIST CAISI evaluation (May 2026) placed it ~8 months behind the US frontier. V4-Pro-0813 (MIT, August 13) bundles a DSpark speculative-decoding module in the checkpoint. API since August 16: peak $1.32 /$ 3.96 per 1M (cache hit $0.044), off-peak half. DeepSeek announced on September 10 that v4-pro traffic would route to V4.1-Flash from September 14, then reversed that on September 11 "in response to user demand"; V4.1-Pro had not launched as of October 1.
DeepSeek V4 Flash	284B total / 13B active (MoE)	1M	Superseded	Replaced by V4.1-Flash. The legacy names `deepseek-v4-flash` and `deepseek-v4-flash-vision-exp` now route to V4.1-Flash at its rates. It was $0.14 /$ 0.28 before the August 16 repricing.
DeepSeek-V3.2	671B (MoE)	128K	Previous gen	General-purpose; 98% cache-hit discount ( $0.28/$ 0.42 per 1M base at the time). Superseded by the V4 line for new builds.
DeepSeek-V3	671B (MoE, 37B active)	128K	Previous gen	GPT-4o level at a fraction of training cost; open weights.
DeepSeek-R1	671B (MoE)	128K	Reasoning	Matched o1 on math/code; first open-source reasoning model.
DeepSeek-R1-Distill	7B to 70B	-	Reasoning	Distilled to smaller models; cost-efficient reasoning.

Key October 2026 context: DeepSeek is no longer the automatic cheap answer, and its prices now depend on the clock. The August 16 repricing moved V4 to peak and off-peak billing: peak is 01:00-04:00 and 06:00-10:00 UTC on weekdays, excluding Chinese public holidays, and everything else, including weekends, is half price. V4.1-Flash then cut Flash prices on September 10, but at peak its $1.20 output only matches GPT-5.6 Luna and sits well above GPT-6 Luna's$ 0.50. Any cost model still using the pre-August flat $0.14 /$ 0.28 or $0.435 /$ 0.87 is off by roughly 2x to 4x. Cache-hit discounts remain steep (about 98% on V4.1-Flash), so cache-friendly batch work scheduled off-peak is where DeepSeek still wins. The V4-Pro reroute-then-reversal within a day is the operational lesson: pin dated builds and alert when a "Pro" endpoint starts serving something else.

Moonshot Kimi Family

Model	Parameters	Context	Notes
Kimi K3	2.8T total / 104B active (MoE)	1M	July 16, 2026 (open weights July 27). Largest open-weight model to date. Always-on thinking, multimodal. $3 /$ 15 per 1M ( $0.30 cached input). License: a custom Kimi K3 License, not Modified MIT. A Model-as-a-Service operator whose revenue (with affiliates) exceeds US$ 20M over 12 months needs a separate agreement before commercial use, and products above 100M MAU or US$20M monthly revenue must display "Kimi K3"; internal use and Moonshot's certified inference partners are exempt. AA v4.3.2: 44 (the 57.1 quoted at launch was on an earlier index version). First open model to top WebDev Arena at launch.
Kimi K2.6	1T total / 32B active (MoE)	-	Released April 20, 2026. Modified MIT license. Native video input; Agent Swarm scaling to 300 sub-agents and 4,000 coordinated steps. Tied GPT-5.5 on SWE-Bench Pro (58.6%); SWE-bench Verified ~80.2%.
Kimi K2.7 Code	1T total / 32B active (MoE)	256K	June 12, 2026. Coding-focused build on K2.6 (Modified MIT) with a MoonViT vision encoder. Reports about +21.8% over K2.6 on Moonshot's own Kimi Code Bench v2 with roughly 30% fewer thinking tokens (vendor benchmark). API about $0.95 /$ 4.00 per 1M. Superseded by K3 as Moonshot's flagship.
Kimi K2-Thinking-0905	-	-	First model to hit 100% on AIME 2025 (reasoning variant).

Best for: Long-horizon agent workloads, video understanding, open-weight agent stack alternative to closed frontier.

K3 as a pretraining substitute: Cognition's SWE-2 (September 10, Devin-only, no weights) is post-trained from K3, and Fireworks' Ember-1 (September 23, research preview at K3's $3 /$ 15) is built on it, with about 40% fewer tokens than K3 (vendor-reported). Western product labs now buy an open base and compete on RL and post-training, which turns the base model's license into a supply-chain term: K3's MaaS clause is written for exactly this kind of hosted API. Moonshot published no new open model between K3 and October 1.

Alibaba Qwen Family

Model	Parameters	License	Notes
Qwen3.8-Flash-Next	125B total / 6B active MoE, plus a 51B N-gram memory and a 4B MTP module	Qwen Community License 1.0	August 26, 2026. An early preview of the Qwen4 architecture: Gated DeltaNet linear attention in three of every four layers plus Qwen Sparse Attention, and an N-gram lookup table that can live in host RAM rather than HBM. 262K native, 1M with YaRN; text, image, and video in. AA v4.3.2: 40. License: any MaaS or AI coding/office-assistant business needs a separate license from Qwen before commercial use, with no revenue floor; internal use is exempt. Hosted as `qwen3.8-flash` on Qwen Cloud at $0.15 /$ 0.47 per 1M.
Qwen3.8-Max	2.4T total / 95B active MoE	Qwen3.8-Max License (gated)	August 12, 2026; open checkpoint Qwen3.8-2.4T-A95B. 262K context native, extensible to ~1,010,000. Open weights under a bespoke license, not Apache: attribution required above 100M MAU or $20M monthly revenue, and a separate license from Qwen is required for Model-as-a-Service or AI Work Assistant (coding or office) businesses above US$ 50M aggregate revenue over 12 months. The open checkpoint is text-input-only; the API version is multimodal. AA v4.3: 40.
Qwen3.8-Max-0902	Same 2.4T / 95B	API only	September 2, 2026 post-training refresh with no weights; US $2 /$ 6 per 1M on Alibaba Model Studio (international), the same as `qwen3.8-max`. The open checkpoint stays at the August 12 build, so self-hosters fall behind the API model within weeks: "open snapshot, closed refresh".
Qwen3.8-27B	27B dense	Apache 2.0	August 14, 2026. The more permissive and more modality-complete artifact: accepts image and video input where the open Max checkpoint does not. 262K context, extensible to ~1M. Dense rather than MoE, which makes it the practical single-GPU option. AA v4.3.2: 34. Base for Ternary Bonsai 2; Groq moved its Qwen tier to it on September 14.
Qwen 3.6 Max-Preview	~1T MoE	Commercial preview	Released ~April 20-27, 2026. 262K context. Topped six coding benchmarks per Alibaba.
Qwen 3.6-Plus	-	-	Released April 2, 2026. Enhanced coding.
Qwen 3.6-35B-A3B	35B / 3B active MoE	Apache 2.0	Released April 16, 2026. Open-weight workhorse.
Qwen2.5-Coder-32B	32B	Apache 2.0	Previous-generation open coding leader.
Qwen2.5-72B	72B	Apache 2.0	Previous-generation multilingual leader.
Qwen2.5-7B	7B	Apache 2.0	Efficient self-hosted option.

Within one family, the newest release (Flash-Next) carries the strictest license and the smaller 27B the most permissive one. Read the LICENSE file for each checkpoint, not the family name.

Mistral Family

Model	Parameters	Context	Notes
Mistral Medium 3.5	128B dense	256K	Released April 29, 2026. Merges Magistral (reasoning) + Pixtral (vision) + Devstral 2 (coding) into one model. 77.6% on SWE-Bench Verified. $1.50 /$ 7.50 per 1M. License: Modified MIT that withdraws all rights from organizations with more than US$20M global monthly revenue (they need a commercial license or Mistral's hosted API), including derivatives. AA v4.3.2: 14. Replaced Medium 3.1, retired August 31.
Voxtral TTS	4B open-weights	streaming	March 23, 2026 release, CC BY-NC 4.0 (non-commercial). 70ms latency, 9 languages, 3-second voice cloning.
Mistral Large 3	675B (MoE, 41B active)	256K	Sparse MoE; parity with best open-weight models at launch; #2 OSS non-reasoning on LMArena at the time. $0.50 /$ 1.50 per 1M on the API.
Mistral Small 4	-	256K	Hybrid instruct/reasoning/coding; released March 2026. $0.15 /$ 0.60 per 1M.
Mistral 3 (14B/8B/3B)	3B to 14B	-	Unified family: multilingual, multimodal, Apache 2.0.
Mixtral 8x22B	141B (MoE)	-	Previous gen; still viable for throughput.

No new Mistral open-weight LLM shipped between August 15 and October 1; the latest open releases are Leanstral 1.5 and Shieldstral 1.0. La Plateforme now also serves Z.ai's GLM 5.3 (GA September 28; GLM 5.2 retires there October 31), so Europe's leading "sovereign" provider resells a Chinese open-weight model. Provenance and hosting jurisdiction are separate questions.

Google Gemma Family

Model	Parameters	Context	License	Notes
Gemma 4 (31B dense)	31B	256K	Apache 2.0	Released April 2, 2026. 140+ languages; native vision/audio; function calling.
Gemma 4 (26B-A4B MoE)	26B / 4B active	256K	Apache 2.0	Sparse MoE variant.
Gemma 4 12B (Unified)	11.95B dense	256K	Apache 2.0	June 3, 2026. Encoder-free multimodal: raw image patches and audio waveforms are projected into the LLM embedding space through lightweight linear layers; text, image, audio, and video (as frames) in. Official QAT int4 builds. The obvious laptop-class multimodal default.
Gemma 4 E4B	8B	256K	Apache 2.0	Edge-suitable.
Gemma 4 E2B	5.1B / 2.3B active	256K	Apache 2.0	Smallest variant; mobile/embedded.
DiffusionGemma (26B-A4B MoE)	26B / ~4B active	256K	Apache 2.0	June 10, 2026. Google DeepMind's first open-weight text-diffusion model; denoises blocks of tokens in parallel for roughly 4x faster generation (1000+ tokens/sec on one H100). Lower quality than standard Gemma 4; aimed at low-latency and in-line editing.

Google released no new Gemma model between August 15 and October 1, and Gemma 5 has not been announced.

Zhipu / Z.ai GLM Family

Model	Parameters	Context	License	Notes
GLM-5.3	744B total / 40B active (MoE), same base as GLM-5.2	1M	GLM-5.3 License (MIT plus a security-review clause)	Coding Plan August 14, 2026; general API August 18; open weights August 28, the date Z.ai had targeted. The entire gain comes from extended post-training rather than a new pretraining run. The license is MIT except that a Model-as-a-Service operator with more than US $10B aggregate revenue over 12 months must pass a Z.ai security review before commercial use, so almost everyone gets effectively MIT terms. API$ 1.40 / $0.26 cached /$ 4.40 per 1M. AA v4.3.2: 45. NIST CAISI (September 17) called it the most cyber-capable open-weight model released to date, about four months behind the US frontier.
GLM-5.3-Flash	320B total / 18B active (MoE)	1M (128K max output)	MIT	August 26, 2026, with none of GLM-5.3's review clause. First natively multimodal GLM-5 model, built for GUI observation and interaction. Hybrid linear and sparse attention (Z.ai: 3.01x less attention compute and a 4.44x smaller KV cache than GLM-5.3). $0.15 /$ 0.50 per 1M, about 1/9 of GLM-5.3. AA v4.3.2: 42.
GLM-5.2	744B total / 40B active (MoE)	1M	MIT	Coding-plan access June 13, 2026; open weights June 16-17. Built for long-horizon agentic coding and tool use. Reports SWE-Bench Pro 62.1 (ahead of GPT-5.5 at 58.6 on that benchmark); vendor-reported. API roughly $1.40 /$ 4.40 per 1M. Superseded by GLM-5.3.

Best for: Open-weight agentic coding and long-horizon tool use where a 1M context and a permissive license matter. GLM-5.3-Flash is the new default self-host candidate in the 18B-active class. Verify benchmark claims on independent leaderboards.

A safety window is not a safety guarantee. Within days of the GLM-5.3 weight drop, Hugging Face carried community "uncensored" and abliterated derivatives, including a GLM-5.3-Flash build with roughly 220K monthly downloads. A staged release delays open access; it does not stop safety training from being stripped afterward.

Thinking Machines Inkling - July 2026

Model	Parameters	Context	License	Notes
Inkling	975B total / 41B active (MoE)	1M (64K or 256K via the lab's Tinker API)	Apache 2.0	Released July 15, 2026: Thinking Machines Lab's first public model, pretrained on 45T tokens of text, image, audio, and video. SWE-Bench Verified 77.6%; the leading US open-weights model per Artificial Analysis at launch, with safety scores the lab reports as aligning with frontier models. NVFP4 checkpoint optimized for NVIDIA Blackwell. AA v4.3.2: 25.
Inkling-Small	~266B total / 12B active	-	Apache 2.0	Full weights released July 30, 2026 (post-trained beyond the earlier preview checkpoint), plus an NVFP4 build. Text, image, and audio in. Vendor-reported SWE-bench Verified 80.2%, beating its larger sibling. AA v4.3.2: 26, the top US open-weight model.

Why it matters: A brand-new US lab shipping the leading American open-weights family, positioned explicitly as a fine-tuning foundation. Kimi K3 and Inkling landing in the same July week made it the strongest open-weights month on record at the time. By October both Inkling models trail MBZUAI's K2 Horizon 375B (31) and sit well behind the Chinese leaders (44 to 46), and no new Thinking Machines model shipped between August 15 and October 1.

Meta Muse Spark (Closed Weights) - April 2026 STRATEGIC SHIFT

Attribute	Value
License	Closed weights - first proprietary model from Meta Superintelligence Labs
Capabilities	Multimodal reasoning with Instant / Thinking / Contemplating modes
Released	April 8, 2026

Strategic significance: Meta's first non-open model since the original Llama era. Signals that frontier-quality work may require a closed-development feedback loop. Meta paused the Llama 4 Behemoth release at the same time amid capability concerns, and Behemoth has never shipped. The open-vs-closed equilibrium became two-tier: frontier closed models lead, and open weights catch up via distillation, RL, and ecosystem iteration. The lead was commonly put at 6 to 12 months in April; by September NIST CAISI estimated GLM-5.3 at about four months behind the US frontier on cyber, while AA v4.3.2 still shows a 12-point gap (58 vs 46).

July 2026 update: Muse Spark 1.1 shipped July 9 alongside the public preview of the Meta Model API, Meta's first self-serve paid API: OpenAI-compatible, $1.25 /$ 4.25 per 1M, roughly a quarter of rival flagship rates. Vendor-reported benchmarks lead on scaled tool use (MCP Atlas 88.1) and professional tool use (JobBench 54.7). Meta charging for API access completes the pivot away from open-weight Llama; there is no Llama 5, and Behemoth remains shelved.

August 2026 update: Muse Glimmer (August 10) is Meta's first open-weight release since Llama 4 and, at Apache 2.0, its most permissive license ever for an open model. It is a 30B dense multimodal model aimed at always-on local agent work (local coding agents, function calling, LLM-as-a-judge), with a 131,072-token context; quantized to 4-bit it fits under 20GB and runs on a single 24GB consumer GPU. Muse Spark 1.2 and Muse Code, a terminal coding agent, shipped alongside it, notable for a muse-spark-1.2-contributor tier priced at $0.10 /$ 0.20 per 1M (against $1.25 /$ 4.25 standard) in exchange for permission to train on your prompts and completions. Data-for-discount as an explicit, published API tier is new, and worth a policy decision before anyone enables it.

September 2026 update: Muse Spark 1.3 (muse-spark-1.3, September 2) keeps the $1.25 /$ 4.25 price ( $0.15 cached) and adds a 1M-token context with text, image, video, and PDF input; the contributor tier (about$ 0.10 / $0.20, traffic may be used by Meta) continues. The `max` tier launched in limited preview pending safety testing and is now listed on Standard. AA v4.3.2 scores it 48 (max). It is more verbose than 1.2: The Decoder put cost per task at$ 0.55 versus $0.40 despite flat per-token prices, a clean example of why cost per task, not cost per token, should drive selection. Meta also launched Muse, a consumer personal agent that can browse and fill out forms (September 8). Muse Glimmer is still Meta's latest open model.

Xiaomi MiMo-V2.6 - September 2026 NEW

Model	Parameters	Context	License	Notes
MiMo-V2.6-Pro	1.02T total / 42B active (MoE)	1M	MIT	September 21, 2026, ungated weights. Omni-modal input (text, image, video, audio). 60 sliding-window plus 10 global attention layers and a 5-layer MTP drafter that predicts 7 tokens per pass. AA v4.3.2: 46, #1 among open-weight models. Vendor-reported Terminal-Bench 2.1 89.9, DeepSWE v1.1 71.9. OpenRouter $0.435 /$ 0.87 per 1M. Xiaomi's SGLang recipe serves it across two nodes with 16-way tensor parallelism.
MiMo-V2.6-Flash	309B total / 15B active (MoE)	1M	MIT	Same date and modalities; fits one 8-GPU node. OpenRouter $0.14 /$ 0.28 per 1M.

Why it matters: a phone maker now leads the open-weight leaderboard, and its OpenRouter prices are exactly DeepSeek's old V4 Pro and V4 Flash list prices, so it took over the price point DeepSeek vacated in August.

Other Open-Weight Families to Know (October 2026)

Model	Released	Size	License	Notes
Tencent Hy4 preview	Aug 28, 2026	770B / 49B active MoE	Apache 2.0 (standard text; no use-policy appendix, revenue gate, or territorial carve-out)	1M context; Gated DeepSeek Sparse Attention. Vendor-reported SWE-Bench Pro 65.7. TokenHub $0.834 /$ 2.501 per 1M. Labeled a preview with known issues such as over-verification. Notable from a lab whose April Hy3 preview license excluded the EU, UK, and South Korea
MBZUAI K2 Horizon	Sep 3, 2026	Six models, 0.9B to 375B-A23B	Apache 2.0, with training data, code, intermediate checkpoints, and logs	512K context (0.9B: 128K). The 375B-A23B scores AA 31, above Inkling and Nemotron 3 Ultra but behind South Korea's Motif 3 (34). The best fully reproducible family for provenance and contamination audits
IBM Granite 4.2	Aug 25, 2026	3B, 8B, 30B dense	Apache 2.0	128K native, extendable to 512K; full, non-thinking, and low-effort modes; official FP8, MXFP4, NVFP4, GGUF, and MLX builds. The common answer to "US-origin, Apache, one GPU"
NVIDIA Nemotron 3 Ultra	Jun 4, 2026	550B / 55B active	OpenMDW-1.1	Not 500B. AA 23. A Labs competitive-coding variant (Sep 3) reached IOI 2026 gold-level scores in an unofficial run using traces distilled from Z.ai's GLM-5.2, so distillation lineage now crosses borders both ways. Nemotron 4 is reportedly still in training; NVIDIA has given no date
MiniMax-M3	Jun 1, 2026	~428B / 23B active	minimax-community	1M context with MiniMax Sparse Attention; AA 29. M3.1 has no public weights or pay-as-you-go API, only a subscription-plan Flash preview
Ant Group Ling-3.0-flash-VL / LLaDA2.2-mini	Sep 4 / Sep 5, 2026	124B / 5.5B active; 16B / 1.4B active	MIT; Apache 2.0	A GUI-agent vision-language MoE, and a diffusion LM with DELETE and INSERT edit tokens and a 128K context
StepFun Step 5 Preview	Sep 20, 2026	Not confirmed	Weights not released	API-only at $1 /$ 2.70 per 1M; AA 44. Verbose: about 160M output tokens on AA's suite vs an 81M median, which eats much of its per-token advantage

The Open-Weight Licensing Split (August to October 2026)

August was the month open weights stopped meaning one thing, and September made the pattern clear: gated licenses converge on one trigger, whether you sell inference or an AI assistant. The license is as much a design input as the benchmark:

Model	Released	Size	License posture	What the license actually does
Qwen3.8-Flash-Next (Alibaba)	Aug 26	125B / 6B active	Qwen Community License 1.0	Any MaaS or AI coding/office-assistant business needs a separate license before commercial use, with no revenue floor; internal use exempt; attribution above 100M MAU or US$20M monthly revenue. Stricter than the flagship's license
Qwen3.8-Max (Alibaba)	Aug 12	2.4T / 95B active	Bespoke, commercially gated	Free use, modification, and resale, but attribution is required above 100M MAU or $20M monthly revenue, and a separate license from Qwen is required to run a Model-as-a-Service or AI Work Assistant (coding or office productivity) business when aggregate revenue with affiliates exceeds US$ 50M over any 12 consecutive months
Kimi K3 (Moonshot)	Jul 27 (weights)	2.8T / 104B active	Kimi K3 License	Separate agreement for MaaS operators above US $20M revenue over 12 months; attribution above 100M MAU or US$ 20M monthly revenue; internal use and certified partners exempt
Mistral Medium 3.5	Apr 29	128B dense	Modified MIT	A blanket revenue cap, not a MaaS trigger: no rights at all above US$20M global monthly revenue
GLM-5.3 (Z.ai)	Aug 28 (weights)	744B / 40B active	GLM-5.3 License (MIT plus review)	Security review before commercial use only for MaaS operators above US$10B aggregate revenue; effectively MIT for almost everyone. Weights shipped on the stated date after a two-week safety window
Qwen3.8-27B (Alibaba)	Aug 14	27B dense	Apache 2.0	No conditions. The smaller sibling is both more permissive and more modality-complete than the flagship
Muse Glimmer (Meta)	Aug 10	30B dense	Apache 2.0	No conditions; Meta's most permissive open license to date
Tencent Hy3	Global Aug 5 (model Jul 6)	295B / 21B active	Apache 2.0	Fully permissive with no geographic carve-outs, reversing the April preview's restrictive license that excluded the EU, UK, and South Korea
Tencent Hy4 preview	Aug 28	770B / 49B active	Apache 2.0	No conditions; verbatim Apache text with no use-policy appendix
MiMo-V2.6-Pro / Flash (Xiaomi)	Sep 21	1.02T / 42B; 309B / 15B	MIT	No conditions
DeepSeek V4.1-Flash	Sep 10	552B backbone	MIT	No conditions
GLM-5.3-Flash (Z.ai)	Aug 26	320B / 18B active	MIT	No conditions; none of GLM-5.3's review clause
K2 Horizon (MBZUAI)	Sep 3	0.9B to 375B	Apache 2.0	No conditions, plus open training data and code
Ling-3.0-flash / tiny (Ant Group)	Aug 5 / Aug 11	124B / 5.1B active; 7.9B / 1.3B active	MIT	No conditions
Nemotron 3.5 Lightning (NVIDIA)	Aug 11	30B / 3B active	OpenMDW-1.1	Linux Foundation license, free for commercial use
Qwen3.8-Max-0902, Step 5 Preview	Sep 2, Sep 20	-	Weights not released	API-only; do not plan around weights until they ship

Three things follow for anyone building on open weights. First, read the license before the model card: the trigger is almost always "are you selling inference or an AI assistant", which is exactly the business many teams (and interview candidates) propose building, and a gated license can make an "open" flagship unusable for it while its smaller sibling is unencumbered. Second, staged release is now a working precedent, and so is the closed refresh: Z.ai shipped GLM-5.3 commercially, held the weights for a two-week safety window, and released them on the date it named, while Alibaba's Qwen3.8-Max-0902 refresh stayed API-only, so do not build a roadmap on weights that have not shipped. Third, mirror what you depend on: NVIDIA agreed to buy Hugging Face for $12.93B (confirmed September 3; no closing date disclosed), so once the deal closes, the main open-weight distribution hub will be owned by the company selling the GPUs those pipelines run on. Keep internal mirrors of the exact weights and license files you deploy.

Small and On-Device Models (October 2026)

Model	Size	Context	License	Notes
Ling-3.0-tiny (Ant Group)	7.9B / 1.3B active	256K	MIT	Genuinely laptop-class agentic model: roughly 8.3 GiB peak memory at 8K context, 86-90 tok/s on an M4 Pro MacBook at FP8
LFM2.5-2.6B (Liquid AI)	2.6B	128K	Open weights	Built for on-device tool calling and multi-step planning rather than chat; under 2.5 GB memory, runs down to a Raspberry Pi
LFM2.5-VL-3B (Liquid AI)	3B	-	Open weights	On-device vision-language model for screen understanding and GUI grounding (August 12)
Shieldstral 1.0 (Mistral)	3B	32K	Apache 2.0	Policy-adaptive multimodal safety classifier (text and image, 12 languages): moderation policies are supplied in natural language at inference time, so a policy change needs no retraining. Runs on one 16GB GPU
Muse Glimmer (Meta)	30B dense	131K	Apache 2.0	Under 20GB at 4-bit; targets always-on local agents
Gemma 4 12B Unified (Google)	11.95B dense	256K	Apache 2.0	Encoder-free multimodal (text, image, audio, video frames); official QAT int4 builds
Granite 4.2 3B / 8B (IBM)	3B, 8B dense	128K (512K extended)	Apache 2.0	Official FP8, MXFP4, NVFP4, GGUF, and MLX builds
K2 Horizon 0.9B / 3.7B / 7B (MBZUAI)	0.9B to 7B	128K (0.9B), 512K	Apache 2.0	MBZUAI says the 0.9B is small enough for a watch; GGUF builds at launch
Ternary Bonsai 2 27B (PrismML)	27.36B at 1.72 bits/weight	262K	Apache 2.0	Derived from Qwen3.8-27B. A 5.95 GB file vs ~54 GB at FP16; PrismML reports 98.2% of FP16 quality across 14 thinking-mode benchmarks. Needs PrismML's llama.cpp and MLX forks (stock llama.cpp cannot load it)
Needle 3 (Cactus Compute)	121M	-	Apache 2.0	On-device tool calling, structured extraction, and embeddings in an 8-29 MB file at about 2 bits per weight
Apple Foundation Models 3 (on-device)	3B dense; 20B sparse activating 1-4B	-	OS-provided	The 20B Core Advanced model keeps its full weights in NAND flash and loads selected experts into DRAM; ships in the iOS 27 cycle

The shift: the small-model tier stopped competing on chat quality and started competing on tool calling, planning, and screen grounding, which are the capabilities a local agent actually needs. September added two hardware-relevant patterns: ternary-native weights that put a 27B reasoner in about 6 GB, and flash-resident MoE that keeps most of a model out of DRAM. Both change the sizing math in On-Device and Edge Deployment.

Specialized Models

Coding (October 2026)

SWE-bench Verified and the public SWE-Bench Pro split are saturated (Opus 5 scores 99.4% on the Pro v2 public split versus 81.6% on the private set), so cite Terminal-Bench 4.0 and SWE-Bench Pro v2 private, with effort and runner.

Model	Role	Evidence
Claude Opus 5.5	Default agentic coding pick at $4/$ 20	Terminal-Bench 4.0 66.4% at xhigh (Anthropic-reported); Claude Code's default Opus since 2.1.280
Claude Sonnet 5.5	Workhorse coding at $2/$ 10	Terminal-Bench 4.0 70.6% (Anthropic-reported); Claude Code's Sonnet default since 2.1.284
GPT-6 Astra	Ceiling; tops the Terminal-Bench 4.0 leaderboard	58.18% at max (tbench.ai, September 21); about a third of GPT-5.6 Sol's tokens per coding task (AA)
Claude Fable 5.1	Ceiling on the Claude side	Terminal-Bench 4.0 57.88% at max (leaderboard); 55.8% in Anthropic's run
GPT-6.1 Sol	OpenAI's $2/$ 10 coding default	Codex CLI default since rust-v0.159.1 (September 29)
Gemini 3.8 Flash	Cheapest near-frontier coder	DeepSWE v1.1 tie near 74% at $2.36 per task (vs Astra xhigh$ 4.43, Opus 5 max $11.84); SWE-Bench Pro v2 private 211/272
Claude Opus 5	Prior Claude flagship	SWE-Bench Pro v2 private 222/272 (81.6%), the top private score Scale published; Terminal-Bench 4.0 53.94% at xhigh
Grok 4.7	Cursor-native option	Terminal-Bench 4.0 37.58% at xhigh (leaderboard); on all Cursor plans
Kimi K3	Largest open model	SWE-Bench Pro v2 private 214/272; custom license gates MaaS use
GLM-5.3 / GLM-5.3-Flash	Open agentic coding	GLM-5.3 SWE-Bench Pro v2 private 211/272; Flash is MIT at $0.15/$ 0.50
MiMo-V2.6-Pro	Top open model overall	AA v4.3.2 46; MIT

Reasoning & Math

Model	Approach	Best For
GPT-6 Astra	Effort `low` to `max`, no `none`	ARC-AGI-3 62.7% on ARC Prize's Standard harness (max, $26,098) and 99.9% on the Provider Adapter harness (high,$ 18,817); HLE-Diamond without tools 59.9% vs Opus 5.5 54.6%
Claude Opus 5.5 / Fable 5.1	Always-on adaptive thinking	Anthropic's HLE with tools: Opus 5.5 67.7% vs Astra 57.2% (vendor-reported). On HLE-Diamond with tools the order flips (Astra 82.9%, Opus 5.5 73.9%), so pick the set that matches your task
Gemini 3.1 Pro Deep Think	Sustained chain-of-thought	Scientific reasoning; Google's top Pro-tier option until Argon reaches the API
DeepSeek V4.1-Flash	Effort dial 1-100	Open (MIT) reasoning at peak $0.30/$ 1.20
DeepSeek-R1	RL-based thinking	Historical open reasoning baseline
Grok 4.3 (DeepSearch)	Web-grounded reasoning	Research tasks needing live information

GPQA Diamond is saturated (Astra about 96%), so it no longer separates frontier models.

Long Context (1M+)

Model	Window	Pricing past the threshold
Llama 4 Scout	10M	Self-hosted; industry-leading open-weight window
Claude Opus 5.5 / Sonnet 5.5 / Fable 5.1	1M	Flat to 1M on Claude 4.6 and later, including fast mode
GPT-6 Astra / Sol / Luna, GPT-6.1 Sol	1.05M	The whole request bills at long-context rates (2x input, 1.5x output) above 272K input
Gemini 3.8 Flash / 3.1 Pro Preview	~1M input	3.1 Pro Preview moves to $4/$ 18 above 200K
Grok 4.7	500K	Every token doubled once the prompt reaches 200K
MiMo-V2.6, GLM-5.3, DeepSeek V4.1-Flash, Hy4 preview	1M	Open weights; KV cost per token varies by orders of magnitude across architectures
Gemini 4 Argon	1M output limit (announced)	Not GA

Embedding Models

API Embedding Models (October 2026)

Model	Max input	Cost/1M	Notes
OpenAI text-embedding-3-large	8,192	$0.13	Unchanged; OpenAI shipped no new embedding model in 2026, and there is no "text-embedding-4"
OpenAI text-embedding-3-small	8,192	$0.02	Cheapest mainstream API option
Voyage voyage-4-large / voyage-4 / voyage-4-lite	32K	$0.12 /$ 0.06 / $0.02	January 15, 2026; one shared embedding space across the three sizes
Voyage voyage-context-4, voyage-code-4	-	$0.12	voyage-code-4 +27.54% over voyage-code-3 (vendor-reported)
Cohere embed-v5.0-pro / embed-v5.0-fast	128K	$0.12 /$ 0.08	September 30, 2026; shared embedding space
Google gemini-embedding-2	-	See Google's pricing page	Stable April 2026; multimodal. Replaces text-embedding-004 (shut down January 14, 2026)

The MTEB scores this table used to carry were 2024-25 numbers that do not compare with current models. Prefer RTEB (retrieval-focused, with private sets) plus an eval on your own corpus.

Open Source Embedding Models

Model	Dimensions	Max Tokens	MTEB (legacy)	Notes
Nemotron 3 Embed 8B / 1B (NVIDIA)	-	-	-	July 16, 2026, OpenMDW-1.1; RTEB Multilingual #1 at 78.5 (vendor-reported)
pplx-embed-v2-context-9b-preview (Perplexity)	-	-	-	September 25, 2026 preview; do not mix its vectors with later releases
BGE-large-en-v1.5	1024	512	63.9	Instruction-tuned
E5-mistral-7b-instruct	4096	32768	66.6	Strong with instructions
Nomic-embed-text-v1.5	768	8192	62.3	Long context, open
GTE-Qwen2-7B	3584	32K	72.1	Strong 2024 open baseline

Embedding Selection Guide

Requirement	Recommended	Why
Best quality	voyage-4-large or embed-v5.0-pro, validated on your corpus	Current flagship APIs; benchmark on RTEB, not MTEB alone
Cost-efficient	voyage-4-lite or text-embedding-3-small	$0.02/1M
Self-hosted	Nemotron 3 Embed 8B	Open (OpenMDW-1.1); RTEB Multilingual leader (vendor-reported)
Long documents	embed-v5.0 (128K) or voyage-4 (32K)	Long input windows
Multilingual	Nemotron 3 Embed	RTEB Multilingual leader (vendor-reported)
Multimodal	gemini-embedding-2	Multimodal input in one model

See Embeddings and Vector Spaces for how to evaluate them.

Model Selection Framework

Decision Tree

What is your primary constraint?

├── Cost → Use smaller model, consider open source
│   ├── Very cost sensitive → GPT-6 Luna ($0.10/$0.50), Gemini 3.8 Flash (budget on the 2027 price), DeepSeek V4.1-Flash (off-peak), Claude Haiku 4.5
│   └── Moderate budget → Claude Sonnet 5.5, GPT-6 Sol or GPT-6.1 Sol ($2/$10), Muse Spark 1.3
│
├── Quality + Reasoning → Use frontier models
│   ├── Highest reasoning → Claude Fable 5.1, GPT-6 Astra (xhigh/max), Claude Opus 5.5 at high effort
│   └── Coding + reasoning → Claude Opus 5.5 (start here), Claude Sonnet 5.5, GPT-6.1 Sol
│
├── Latency → Use fast models or paid speed tiers
│   ├── Interactive and cheap → GPT-6 Luna, Gemini 3.8 Flash at low thinking, Claude Haiku 4.5
│   └── Frontier quality, faster → Opus 5.5 fast mode ($8/$40), GPT-6 Astra Fast (2x) or Ultrafast (6x)
│
├── Self-hosting → Use open models (read the license first)
│   ├── Maximum capability → MiMo-V2.6-Pro (MIT), GLM-5.3 (MIT plus review above $10B), Kimi K3 (MaaS gate above $20M)
│   ├── Good balance → GLM-5.3-Flash (MIT), Tencent Hy4 preview (Apache 2.0), Qwen3.8-27B (Apache 2.0)
│   └── Edge/mobile → Gemma 4 12B or E4B, Granite 4.2 3B/8B, K2 Horizon 0.9B-7B
│
└── Privacy → Self-host, ZDR, or residency-pinned endpoints
    └── Opus 5.5 and Sonnet 5.5 support ZDR; Fable 5.1 needs authorization; residency costs ~10% more

Semantic Routing

Static decision trees are being replaced by Semantic Routers:

How it works: A small, fast embedding model embeds the query. If it matches a "known easy" cluster, route to a cheap model (GPT-6 Luna, Gemini 3.8 Flash, DeepSeek V4.1-Flash off-peak). If it hits an "agentic/logic" cluster, route to Claude Opus 5.5 or GPT-6.1 Sol at higher effort, and reserve Fable 5.1 or GPT-6 Astra for what those still fail.
Benefit: Automates cost-optimization without hardcoded rules.
Implementation: Tools like semantic-router (Python), vLLM Semantic Router, or custom Weaviate/Pinecone classifiers.

What September 2026 changed for routers:

Prompt length is a price input. OpenAI bills the whole request at long-context rates above 272K input, xAI doubles every token once a prompt reaches 200K, and Anthropic stays flat to 1M. Route long prompts by price, not just by window size.
Speed is a tier, not a model. OpenAI Fast (2x) and Ultrafast (6x on Astra), Claude fast mode ( $8/$ 40 on Opus 5.5), and Gemini Priority (about 1.8x) mean a latency SLO maps to a tier per route.
Conversation state is model-bound on Claude. Fallback chains that replay a Claude 5.5 or Fable 5.1 conversation to another model lose the reasoning blocks, and history rewrites before a thinking block fail with a 400. Fall back at turn boundaries and keep history append-only.
Safety refusals are routing events. Fable 5.1 refusals come back as HTTP 200 with stop_reason: "refusal"; handle them with fallbacks: "default" or your own retry, and remember that benchmark scores labeled "with fallback" already include that routing.
IDs do not pin behavior. GPT-6 Sol and Luna changed behavior under the same ID on September 25; DeepSeek announced and then reversed a Pro-to-Flash reroute within a day; retired xAI slugs redirect to grok-4.3 at its rates; Codex CLI and Claude Code swapped default models in point releases. Verify the served model, alert on changes, and rerun evals on vendor changelog events.

Sovereign AI and Data Residency

The 2026 Regulatory Reality: Enterprises must comply with GDPR (EU), DPDPA (India), Saudi Arabia PDPL, and sectoral rules. "Sovereign AI" is now a product category.

Solution	Provider	Use Case
Azure Government/Sovereign	Microsoft	Dedicated infra in 40+ regions; approved for US Gov/EU NIS2
AWS Sovereign Cloud	Amazon	Physically isolated VPCs; GDPR-safe EU regions
Google Distributed Cloud	Google	Air-gapped on-prem Gemini deployment
Residency-pinned API endpoints	Anthropic (`inference_geo: "us"`, 1.1x on Claude 4.6 and later), OpenAI (regional endpoints, +10% for models released from March 5, 2026), Bedrock and Google Cloud regional Claude, Bedrock regional GPT-6 Astra, Mistral EU (each about +10%)	Residency without self-hosting, priced as a published line item
Private Llama 4 / 3.3	Meta (self-host)	Maximum data sovereignty; open weights (Llama 4 MoE or 3.3 dense)
Permissive open weights (self-host)	Z.ai GLM-5.3-Flash, Tencent Hy4 preview, Xiaomi MiMo-V2.6, IBM Granite 4.2	MIT or Apache 2.0 weights; no data leaves your infra
DeepSeek (self-host)	DeepSeek (MIT weights)	Open weights; no data leaves your infra
Mistral Large 3 (self-host)	Mistral (Apache 2.0)	675B MoE; open weights; strong multilingual

Tradeoff: Dedicated sovereign clouds are priced by contract and usually carry a premium over standard regions, while per-request residency on the major APIs is now a published ~10% premium. Price residency per tenant and route only residency-bound traffic to pinned geos. Two newer wrinkles matter in regulated designs. Mistral's EU platform now serves Z.ai's GLM 5.3, so model provenance and hosting jurisdiction are separate questions. And frontier zero data retention now comes with monitoring options: Anthropic's Enterprise Frontier Safeguards (announced September 1, rolling out in phases later this fall) let customers opt in to keeping misuse-monitoring data in their own cloud account under customer-managed keys, with fully automated review, and OpenAI began previewing Private Safety Processing with select customers on August 19.

Cost Comparison at Scale (October 2026)

Assume 1M requests/day, 1K input + 500 output tokens, Standard tier list prices, no caching:

Model	Input Cost/Day	Output Cost/Day	Total/Month
GPT-6 Astra	$10,000	$25,000	$1,050,000
Claude Opus 5.5	$4,000	$10,000	$420,000
GPT-5.6 Terra	$2,000	$6,000	$240,000
Gemini 3.1 Pro Preview	$2,000	$6,000	$240,000
Claude Sonnet 5.5	$2,000	$5,000	$210,000
GPT-6 Sol / GPT-6.1 Sol	$2,000	$5,000	$210,000
Gemini 3.8 Flash (from January 1, 2027)	$1,500	$3,750	$157,500
Grok 4.7	$2,000	$3,000	$150,000
Muse Spark 1.3	$1,250	$2,125	$101,250
Gemini 3.8 Flash (intro, through December 31)	$750	$1,875	$78,750
DeepSeek V4.1-Flash (all at peak)	$300	$600	$27,000
GPT-6 Luna	$100	$250	$10,500

How to read it:

Cost per task beats cost per token. Verbose models eat their price advantage (Muse Spark 1.3 at $0.55 vs$ 0.40 per task for 1.2; Step 5 Preview at about twice the median output tokens on AA's suite), and token-efficient ones narrow theirs (Astra uses about a third of GPT-5.6 Sol's tokens per coding task). Measure tokens per task on your workload before trusting this table.
Caching reorders it. With a cached system prompt, cache reads cost 0.1x on most models, 0.05x on Opus 5.5 and GPT-6.1 Sol, and 0.025x on Fable 5.1, so per-model break-even matters more than the headline rate.
Promotions expire. Gemini 3.8 Flash doubles on January 1, 2027, and GPT-5.6 Sol's $4/$ 20 promotion is guaranteed only through at least November 21, 2026.
Self-hosting is priced in GPU-hours, not tokens. On October 1 list prices, an H100 runs $2.69 to$ 3.99 per hour on demand (RunPod community to Lambda) and a B200 $5.98 to$ 6.79 (RunPod community to RunPod secure); see Pricing and Costs for the break-even math.

Capability Comparison

Benchmark Snapshot (October 2026)

Model	AA Index v4.3.2	Terminal-Bench 4.0	SWE-Bench Pro v2 private (of 272)	Notes
Claude Opus 5.5	58 (max, with fallback)	66.4% xhigh (Anthropic)	-	OSWorld 2.1 partial 81.8% (Anthropic)
Claude Sonnet 5.5	56 (max, with fallback)	70.6% (Anthropic)	-	GDPval-AA v2.1 within 2 Elo of Opus 5.5 (Anthropic)
Claude Fable 5.1	53 (max, with fallback)	57.88% max (leaderboard); 55.8% (Anthropic)	-	Cache reads 0.025x
GPT-6 Astra	53 (max)	58.18% max (leaderboard, rank 1)	-	ARC-AGI-3 62.7% Standard vs 99.9% Provider Adapter (ARC Prize)
Gemini 4 Argon (not GA)	53 (high, pre-GA)	-	-	DeepSWE v1.1 77.9% (Google)
GPT-6.1 Sol	52 (max)	-	-	$2/$ 10
Muse Spark 1.3	48 (max)	-	-	Verbose; check cost per task
MiMo-V2.6-Pro (open)	46	-	-	Top open-weight model
Grok 4.7	46 (xhigh)	37.58% xhigh (leaderboard)	-	+2 over Grok 4.6 on AA
GLM-5.3 (open)	45	-	211 (77.6%)	MIT plus MaaS review
Kimi K3 (open)	44	-	214 (78.7%)	Custom license
Gemini 3.8 Flash	41 (high)	-	211 (77.6%)	Intro price through December 31
Claude Opus 5	-	53.94% xhigh (leaderboard)	222 (81.6%)	99.4% on the saturated public split
Inkling (open)	25	-	184 (67.6%)	Inkling-Small scores 26

Sources: Artificial Analysis Intelligence Index v4.3.2 (early October), the tbench.ai Terminal-Bench 4.0 leaderboard (September 21), Scale's SWE-Bench Pro v2 (September 22), ARC Prize (September 3). Vendor figures are labeled with the vendor's name. Rules for reading any table like this:

Index versions do not mix. AA v4.3 (launched September 7) re-baselined everything, so pre-September figures elsewhere in this chapter (Kimi K3 57.1, Grok 4.6 61) cannot sit in the same column.
"With fallback" means a routed system. Claude scores include safeguard fallbacks to Opus 4.8 or Opus 5 when a classifier intervened; pin fallback state the way you pin harness and effort.
Name the effort and the runner. Astra is 58.18% at max on the leaderboard but 57.9% at high in OpenAI's figure; Fable 5.1 is 57.88% on the leaderboard but 55.8% in Anthropic's own run.
Saturated sets stop discriminating. SWE-bench Verified, the public SWE-Bench Pro split, and GPQA Diamond no longer separate frontier models; use private or held-out sets. See Benchmarks and Leaderboards.

Task-Specific Recommendations (October 2026)

Task	Recommended Models	Why
Autonomous Coding Agent	Claude Opus 5.5, Claude Sonnet 5.5, GPT-6.1 Sol	Top vendor-reported Terminal-Bench 4.0 scores (the two Claude models) at $2-$ 4 input; defaults in Claude Code and Codex
Complex Reasoning	Claude Fable 5.1, GPT-6 Astra (xhigh/max), Claude Opus 5.5 at high effort	Ceiling models; Opus 5.5 first, escalate when evals fall short
Agentic Computer Use	Claude Opus 5.5 / Sonnet 5.5 with `computer_toolset_20260801`, GPT-6 Astra, Gemini 3.8 Flash	OSWorld 2.1 partial 81.8% / 80.1% (Anthropic) and 72.6% for Astra on OpenAI's offline set (OpenAI); Gemini's recommended computer-use model
High-Volume API	GPT-6 Luna, Gemini 3.8 Flash, DeepSeek V4.1-Flash (off-peak)	Lowest cost per token in class; budget Gemini on its 2027 price
Long Context RAG	Claude Sonnet 5.5 / Opus 5.5 (flat to 1M), Gemini 3.8 Flash	No long-context surcharge on Claude; keep OpenAI prompts under 272K or price the cliff
Ultra-Long Context	Llama 4 Scout (10M)	Industry-leading 10M context; open weights
Multimodal Real-time	GPT-Live-1, Gemini 3.8 Live	Full-duplex voice with delegated reasoning, or native audio with async tools
Private Production	GLM-5.3-Flash, Qwen3.8-27B, MiMo-V2.6-Flash, IBM Granite 4.2	Permissive licenses; single-node to single-GPU footprints
Open-source Coding	GLM-5.3, MiMo-V2.6-Pro, Kimi K3 (check the license)	Top open scores on AA and SWE-Bench Pro v2 private
Gated Cyber or Bio Work	Claude Mythos 5.1, GPT-5.6-Cyber, Gemini 3.8 Flash Cyber, GPT-Rosalind	Capability-tiered access programs with identity verification
Creative/Chat	GPT-6 Sol, Claude Sonnet 5.5	Strong conversation quality and instruction following at $2/$ 10

Interview Questions

Q: How would you select a model for a production RAG system?

Strong answer: I evaluate across these dimensions:

1. Quality requirements:

Test on representative queries from the actual domain
Measure answer correctness, hallucination rate, citation accuracy

2. Cost analysis:

Monthly cost = requests/day × 30 × (input_tokens × input_rate + output_tokens × output_rate)

Always calculate for the top 2-3 candidates, at the price that applies next quarter (Gemini 3.8 Flash doubles on January 1, 2027), with your measured cache hit rate and tokens per task.

3. Latency requirements:

Tight TTFT budgets: GPT-6 Luna, Gemini 3.8 Flash at low thinking, Claude Haiku 4.5; measure TTFT yourself
If quality is paramount: accept multi-second latency with Claude Opus 5.5 or GPT-6.1 Sol, or buy speed with Opus 5.5 fast mode or OpenAI Fast

4. Operational requirements:

Self-hosting: GLM-5.3-Flash, Qwen3.8-27B, or Llama 4 Scout/Maverick, after a license check
Compliance / data residency: ZDR-eligible models, residency-pinned endpoints (~10% premium), or self-hosted

5. Practical selection:

Start with Claude Sonnet 5.5 or GPT-6 Sol for prototyping
A/B test GPT-6 Luna or Gemini 3.8 Flash for the 80% of easy queries (cost)
Keep Opus 5.5 on hard queries via semantic routing, and keep prompts under 272K on OpenAI models

Q: Explain the tradeoffs between proprietary and open source models.

Strong answer:

Factor	Proprietary (OpenAI, Anthropic, Google)	Open Weights (GLM, Qwen, DeepSeek, MiMo)
Quality	Higher at the top end	12 points behind on AA v4.3.2 (46 vs 58)
Cost	Per-token pricing; $0.10/$ 0.50 at the low end	Compute + ops; cheap only at high utilization
Control	Limited	Full
Privacy	Data goes to provider (ZDR available on many models)	Stays on-prem
Updates	Automatic, sometimes under the same model ID	Manual; you pin the weights
Customization	Limited fine-tuning (OpenAI stops new fine-tuning jobs January 6, 2027)	Full fine-tuning
License	Terms of service	Ranges from MIT/Apache to MaaS-gated custom licenses
Ops overhead	None	Significant

Key insight (October 2026): The gap is real but bounded, and the license is now the bigger variable. On AA v4.3.2 the best open model (MiMo-V2.6-Pro, 46) trails the best closed one (Claude Opus 5.5, 58), and the best US open model (Inkling-Small, 26) trails the best Chinese one by 20 points. Meanwhile GPT-6 Luna at $0.10 /$ 0.50 undercuts most hosted open models, so "open is cheaper" only holds at high utilization on your own GPUs. Choose open weights for control, residency, and fine-tuning, and read the license first: Qwen Community 1.0, Kimi K3, Mistral Medium 3.5, and GLM-5.3 all gate some commercial uses.

Q: How do reasoning controls differ between GPT-6 and the Claude 5.5 generation, and what breaks when you route between them?

Strong answer: Both vendors expose reasoning as an effort dial rather than a toggle, but the contracts differ:

OpenAI GPT-6 family: effort low to max (GPT-6.1 Sol defaults to medium; Astra has no none). Reasoning tokens are billed but not returned. "Pro" is reasoning.mode: "pro" on the GPT-5.6 and GPT-6 families rather than a separate model (GPT-5.4 Pro and GPT-5.5 Pro remain at $30/$ 180). Astra drops temperature, top_p, and logprobs, and its tool calling requires the Responses API.
Claude Opus 5.5 / Sonnet 5.5 / Fable 5.1: adaptive thinking is always on (Sonnet 5.5's floor is between_tools; disabled returns a 400). Default effort differs by model: Opus 5.5 medium, Sonnet 5.5 and Fable 5.1 high on the API. Thinking, including the text between tool calls, arrives in thinking content blocks that are empty at the default display: "omitted" (the display: "updates" beta returns the text). Blocks are bound to the producing model and conversation.
What breaks in a router: replaying a Claude conversation to another model silently drops its reasoning blocks; editing anything before a thinking block (system prompt, tools, earlier turns, even the image bytes behind a URL) returns a 400 for accounts created on or after August 31, 2026; forced tool_choice returns a 400 on the newest Claude models; and an unpinned Opus 5 to 5.5 upgrade quietly moves effort from high to medium.

Production choice: Pin effort per route, keep conversation history append-only, use auto tool choice with strict schemas, and rerun evals at the effort you will actually serve. Log stop_reason, effort, and the served model on every call; visible reasoning is no longer something you can count on from either vendor.

Q: You want to launch a hosted coding assistant on an open-weight model. How do you choose one?

Strong answer: I screen in this order, because the first gate eliminates more candidates than benchmarks do:

License triggers: a hosted coding assistant is exactly what gated licenses target. Qwen3.8-Flash-Next (Qwen Community 1.0) needs a separate license from day one, Kimi K3 needs a separate agreement if the company runs any MaaS business and its total revenue (with affiliates) passes US $20M over 12 months, Mistral Medium 3.5 withdraws all rights above US$ 20M monthly revenue, and GLM-5.3's security review only bites above US$10B. MiMo-V2.6, GLM-5.3-Flash, DeepSeek V4.1-Flash, Tencent Hy4 preview, and Qwen3.8-27B are MIT or Apache.
Quality on like-for-like scales and on my tasks: AA v4.3.2 (MiMo-V2.6-Pro 46, GLM-5.3 45, GLM-5.3-Flash 42) and SWE-Bench Pro v2 private (GLM-5.3 211/272, Kimi K3 214/272), then my own eval on real repositories.
Serving cost by architecture: KV bytes per token differ by orders of magnitude across these models, DeepSeek V4.1-Flash activates 8B parameters at prefill but 16B at decode, and several checkpoints ship their own speculative-decoding heads. Size from the model card, not a generic formula.
Supply chain: mirror the exact weights and LICENSE files, watch for API-only refreshes that leave the open checkpoint behind (Qwen3.8-Max-0902), and treat post-trained derivatives as inheriting their base license.
The closed baseline: compare total cost against GPT-6 Luna or Gemini 3.8 Flash. If my utilization is low, the hosted closed model often wins on cost.

Q: A vendor says its new model beats your current one by 10 points. What do you check before switching?

Strong answer: Six things. Some move a number by a point or two; the harness and index-version effects alone have moved headline numbers by tens of points this year:

Runner: vendor run or public leaderboard? Fable 5.1 is 55.8% on Terminal-Bench 4.0 in Anthropic's run and 57.88% on the leaderboard.
Effort: Astra is 58.18% at max and 57.88% at high on the same leaderboard; compare at the effort you will pay for.
Harness: Astra scores 62.7% on ARC-AGI-3 with ARC Prize's Standard harness and 99.9% with ARC Prize's Provider Adapter harness (which uses OpenAI's own context-management features), same model and test set.
Fallback state: Claude scores "with fallback" belong to a router plus several models.
Index version and saturation: AA v4.3 scores do not compare with earlier versions (Kimi K3 went from 57.1 to 44 by re-baselining alone), and Opus 5 scores 99.4% on the public SWE-Bench Pro split but 81.6% on the private one.
Cost per task: tokens per task differ by about 3x between models of similar quality.

Then I run my own eval at the serving effort, on held-out tasks, with fallback state pinned, and I rerun it after vendor-side changes (GPT-6 Sol's September 25 image fix changed results under the same ID).

References

Anthropic models overview: https://platform.claude.com/docs/en/about-claude/models/overview
Anthropic model deprecations: https://platform.claude.com/docs/en/about-claude/model-deprecations
OpenAI Platform models: https://developers.openai.com/api/docs/models
OpenAI pricing and deprecations: https://developers.openai.com/api/docs/pricing and https://developers.openai.com/api/docs/deprecations
Google AI models and changelog: https://ai.google.dev/gemini-api/docs/models and https://ai.google.dev/gemini-api/docs/changelog
Meta Llama: https://www.llama.com/
DeepSeek pricing: https://api-docs.deepseek.com/quick_start/pricing
xAI Grok: https://docs.x.ai/developers/models
Mistral AI: https://docs.mistral.ai/models/
Artificial Analysis leaderboards: https://artificialanalysis.ai/leaderboards/models
Terminal-Bench 4.0: https://www.tbench.ai/news/terminal-bench-4-0
Scale SWE-Bench Pro v2: https://labs.scale.com/blog/swe-bench-pro-v2
ARC Prize on GPT-6 Astra: https://arcprize.org/blog/astra
Arena (formerly LMArena): https://arena.ai/

Next: Capability Assessment