TUE, NOVEMBER 17, 2026
Independent · In‑Depth · Practitioner‑Tested
✎ General

Every AI Model Released This Year, and What Is Coming Next

Claude Fable 5.2 and Mythos 5.2 shipped on 9 November at $8 per million input tokens, 20% below the 5.1 versions. Meta Muse Ember landed the same day at $0.10. NVIDIA opened GR00T 2 with 100,000 hours of training video. Every 2026 release below, newest first, with what is still only a press release kept separate.

By AIToolsRecap August 6, 2026 9 min read 52469 views
Home › Articles › General › AI Model Release Tracker 2026: Every LLM Launch...

Latest release · 10 November 2026

Gemini 4 Carbon — generally available

Released10 November 2026. Unveiled 28 October with limited access through the Fairwind cybersecurity program.
Price per million$2 in, $10 out through 30 November, then $4 and $20 from 1 December
Context1M tokens
InputsText, images, speech and video. Text out.
WhereGoogle AI Studio paid tier and Android Studio. Google AI Ultra subscribers also get it in Google Antigravity.

Why it matters: at the introductory rate Carbon is a quarter of Claude Fable 5.2 at $8 and $40, and half once the promotion ends. It scores 60 on the Artificial Analysis Intelligence Index against Fable 65, but wins Zapier AutomationBench at 53.88% and Terminal-Bench 4.0 at 65.2%. Cost per finished task runs roughly $7.76 against Fable at about $26.54.

Availability confirmed on the Google blog. Per-token pricing is as reported at launch — Google has not yet added Gemini 4 Carbon to the public AI Studio price list, which still shows only Gemini 4 Argon.

Full launch breakdown · Carbon vs Fable 5.2, benchmark by benchmark

9 November 2026

Claude Fable 5.2 and Claude Mythos 5.2 — released

Claude Fable 5.2 Claude Mythos 5.2
Released 9 November 2026 9 November 2026
Model ID claude-fable-5-2 claude-mythos-5-2
Price / MTok $8 in, $40 out $8 in, $40 out
Fast mode $16 in, $80 out Verified orgs only
Context 1M 1M
Max output 128K 128K
Knowledge cutoff June 2026 June 2026
Availability Open — API, Bedrock, Google Cloud, Microsoft Foundry Cyber or Life Sciences Verification Program
Retirement no sooner than 9 November 2027 9 November 2027

Why it matters: the price went down. Fable 5.1 was $10 / $50, so this is exactly 20% off both sides, with outputs up to 30% faster. Fable 5.2 leads nine of the eleven benchmarks Anthropic published, including FrontierCode 2.0 at 49.1% and FrontierSWE v2 at 69.9%. Mythos is the same model without the cyber and biology safeguards.

Full launch breakdown with every benchmark · Fable 5.2 vs Opus 5.5 vs GPT-6 vs Gemini 4 · Fable vs Mythos: what actually differs

Current as of 7 November 2026
● Newest confirmed release — Kimi K3.1 Pro, 23 October, open weights under a modified MIT
● Newest closed model — Muse Wonder, Meta, 12 October, $4 / $20 per million
● Shipped since our last table — 13 models across five labs
● Still undated — DeepSeek V4-Pro GA, Qwen4, GPT-7, Claude 6

Every model with a confirmed release date, price and model string, most recent first. A model does not enter this table until all three are confirmed from the lab. Anything shipped but not yet priced sits in its own list directly below, and anything without a date sits further down in Targeted, But Not Dated.

Released — Most Recent First

Model Lab Date Price input/M Open weights
Kimi K3.1 Pro Moonshot AI Oct 23 $2.00 / $10.00 ✓ Modified MIT
Muse Wonder Meta Oct 12 $4.00 / $20.00 ✗ Closed
Muse Wonder Full Meta Oct 12 Not on the API ✗ Muse Code only
GPT-6.1 Sol OpenAI Sep 29 $2.00 (cached $0.10) ✗ Closed
Grok 4.7 xAI Sep 21 $2.00 / $6.00 ✗ Closed
Qwen3.8-Flash-Next Alibaba Sep 4 Free ✓ Check licence
K2 Horizon (6 models) Institute of Foundation Models Sep 3 Free ✓ Plus training data
GPT-6 Astra OpenAI Sep 3 Not published ✗ Closed
Claude Fable 5.1 / Mythos 5.1 Anthropic Sep 1 $10 (cache reads $0.25) ✗ Closed
GLM-5.3 Z.ai Aug 14 (weights Aug 28) See Z.ai ✓ Licence unstated
GLM-5.3-Flash Z.ai Aug 26 $0.15 ✓ MIT
Qwen3.8-Max Alibaba Aug 2026 See Alibaba ✓ Custom licence
Claude Sonnet 5 (repriced) Anthropic Aug 31 $3 (was $2) ✗ Closed
Grok 4.6 xAI Aug 2026 $2.00 ✗ Closed
DeepSeek V4 Flash 0731 DeepSeek Jul 31 $0.14 ✓ MIT
Claude Opus 5 Anthropic Jul 24 $5.00 ✗ Closed
Gemini 3.6 Flash Google Jul 21 $1.50 ✗ Closed
Kimi K3 Moonshot AI Jul 16 (weights Jul 27) $3.00 ✓ Modified MIT
GPT-5.6 Luna / Terra / Sol OpenAI Jul 9 $0.20 / $2.50 / $4 promo ✗ Closed
Laguna S 2.1 Poolside Jul 2 $0.10 ✓ OpenMDW
Claude Fable 5 / Sonnet 5 Anthropic Jun 9 / Jun 30 $10 / $3 ✗ Closed
Gemma 4 Google Jun 2026 Free ✓ Apache 2.0
Llama 4 Scout / Maverick Meta Apr 2026 Free ✓ Llama 4 Licence
The Astra story resolved
OpenAI paused Astra on 8 August after evaluations could not rule out a Critical cybersecurity capability. The finding held. On 3 September it shipped anyway, with the cyber capability gated behind a vetted application programme called Daybreak while the rest of the model reaches ordinary subscribers.

Shipped, Pricing Not Yet Confirmed

These have a release date we trust but no price or model string confirmed from the lab itself, so they are not in the table above. Most trackers would list them anyway. We would rather you knew which figures we have checked and which we have not.

  • GPT Image 2.6 Flare and Sunburst — OpenAI, 4 November
  • DeepSeek V4.1 Flash Cyber — DeepSeek, 29 October
  • Kimi K3.1 Pro Highspeed — Moonshot AI, 23 October
  • Qwen3.6 27B — Alibaba, 21 October
  • MiniMax M3.1 Flash — MiniMax, 19 October
  • DeepSeek V4.1 Flash 1017 — DeepSeek, 17 October
  • FLUX 3 Max — Black Forest Labs, 15 October
  • GPT-6.1 Terra — OpenAI, 15 October

Confirmed Dates Ahead

  • 12 November 2026 — OpenAI models leave Cursor. Following the SpaceX acquisition. Anthropic has not said whether Claude follows. Full story
  • Around 21 November 2026 — GPT-5.6 Sol promotional pricing ends. $4 / $20 returns to $5 / $30. Price any migration at the standard rate, not the promotional one.

Recently Resolved

Dates that have passed, and what happened. Items move here from the section above rather than sitting in it after the fact.

  • 23 October — Kimi K3.1 Pro shipped with open weights. 2.8T total, 104B active, 1M context, modified MIT, at moonshotai/Kimi-K3.1-Pro. Moonshot states it reduced the model’s offensive cyber capability before release.
  • 24 October — deepseek-chat and deepseek-reasoner deprecated. Anything still pointing at those IDs is failing. Migrate to deepseek-v4-flash.
  • 12 October — Meta shipped Muse Wonder and Muse Wonder Full, its first frontier-class models. Only Wonder reaches the API at $4 / $20; Wonder Full is restricted to Muse Code subscribers and is not served through the API at all.
  • 29 September — GPT-6.1 Sol. $2.00 input, $10.00 output, $0.10 cached, 1.1M context.
  • 21 September — Grok 4.7 shipped, closing the September window Musk had indicated. Model ID grok-4.7, 500K context, $2.00 / $6.00, cache reads $0.50. What changed in Grok since
  • 31 August — Claude Sonnet 5 repriced from $2 / $10 to $3 / $15 per million, with a tokenizer change adding 10 to 35% more tokens on code. A coding workload can sit 65 to 100% above where it did in August.
  • 31 August — GPT-5.4 and GPT-5.4 mini left Codex for ChatGPT sign-in users. Both remain available via API key, so the same tool behaves differently depending on how you authenticated.
  • 31 August — kimi-k2.5 and moonshot-v1 sunset. Migrate to kimi-k3.

Targeted, But Not Dated

Announced, previewed or reported, with no confirmed date. The distinction matters if you are planning a migration around any of them.

DeepSeek V4-Pro general availability

1.6T flagship, confirmed as coming soon in the July changelog. No date published. A preview is available via the API.

Qwen4 — architecture previewed, model not announced

Alibaba released Qwen3.8-Flash-Next on 4 September specifically to preview the Qwen4 architecture. That usually means a flagship follows within months rather than weeks. What the preview shows.

GPT-7 and Claude 6

Neither appears on any confirmed schedule, and neither lab has announced a date.

OpenAI public listing

Ruled out for 2026 on 12 September. Altman told Fortune he would say not 2026, and declined to confirm 2027. The CFO separately told staff in August that OpenAI will be public in 2027, or sooner if the business continues to inflect. Full story.

Free and Open-Weight Models — Read the Licence First

Open weights and open source have drifted apart
Five open-weight releases since June, five different licences, and only one was Apache. Qwen3.8-Max under custom terms, Kimi K3 and now K3.1 Pro under a modified MIT, GLM-5.3 flagship terms still unstated.
  • Kimi K3.1 Pro (modified MIT, 23 October): 2.8T total, 104B active MoE, 1M context, multimodal. $2.00 / $10.00 hosted, or self-hosted from moonshotai/Kimi-K3.1-Pro. Modified means standard MIT semantics do not carry over — read the file.
  • K2 Horizon (3 September): Six models, 0.9B to 375B. The only 2026 release published with weights, training code, training data, checkpoints and logs. Not a frontier capability claim, a verifiability one. Why the training data matters
  • Qwen3.8-Flash-Next (4 September): Architecture preview for Qwen4. Reported 125B total, around 6B active, plus a 51B component designed for system RAM rather than GPU memory. Check the licence file — Qwen3.8-Max used custom terms.
  • GLM-5.3-Flash (MIT, 26 August): 320B/18B, 1M context, $0.15 / $0.50. Genuinely permissive, unlike the flagship. FP8 checkpoint around 306 GiB, so 4x H200 or 8x H100 minimum.
  • DeepSeek V4 Flash 0731 (MIT, 31 July): $0.14/M or self-hosted. Terminal-Bench 82.7%. Weights at deepseek-ai/DeepSeek-V4-Flash-0731. Full review
  • Kimi K3 (modified MIT, 27 July): 2.8T/104B MoE, SWE Marathon #1. 8x H100 minimum.
  • Laguna S 2.1 (OpenMDW, 2 July): 118B/8B MoE. $0.10/M or self-hosted on a desktop DGX Spark. Terminal-Bench 70.2%. Full review
  • Gemma 4 (Apache 2.0, June): Runs on everyday devices, no API cost. Launch coverage
Six assistants now cost within a dollar of each other
Which means price has stopped being the deciding factor. What decides it is what each bundles, and what happens when you hit a limit — on some tiers you can buy more, on others you stop until the window resets.

Where the Grok plans landed · Claude Pro vs Max · All six assistants compared

How We Track Releases

Updated manually as part of our daily AI news coverage. We source from official lab announcements, API changelogs and verified third-party trackers. A model does not enter the released table until pricing, model string and availability are confirmed from the official source — anything shipped without those sits in Shipped, Pricing Not Yet Confirmed, and anything without a date sits in Targeted, But Not Dated.

Dates that pass move into Recently Resolved with what happened. Confirmed Dates Ahead only ever contains dates that are still ahead, because a tracker that leaves last month in the upcoming section is not worth citing.

Tags
AI NewsGenerative AIBest AI ToolsAI Guide2026

Spot an inaccuracy?

We verify facts before publishing and correct errors promptly. If something in this article is wrong or outdated, let us know.

Report an error →