LLM Release Tracker: Every New Model by Date and Provider

New large language models now arrive every few weeks from each major provider, and teams lose track of which model ids exist, when they shipped and who can use them. This tracker lists every 2026 model release we could confirm on a provider's own release notes, changelog or announcement page, newest first, with the API id and the access status at release. It is written for engineers, platform owners and AI governance teams who approve model changes, and for anyone who wants a dated, sourced record rather than a news summary. Providers are listed alphabetically within each date; order is never a ranking. Last checked: 2026-11-15.

This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider.

Key takeaways

• Release cadence is roughly monthly per major provider. Between July and mid November 2026 the official logs checked show new text models from Anthropic, DeepSeek, Google, Meta, Mistral AI, OpenAI and xAI.
• Restricted-access releases are now common. Anthropic Mythos models and OpenAI cyber and Rosalind models are released to approved, verified or programme participant organisations only, and the tracker states access status the same way for every provider.
• Release date depends on the surface. Google publishes some dates only on its Google Cloud model versions page, and DeepSeek ships updates under a stable alias (deepseek-flash), so record the id and the surface, not just the name.
• Many models are short lived. DeepSeek-V4-Flash went from public beta on 2026-07-31 to retired on 2026-09-10, and Gemini 3.7 Flash was deprecated eight weeks after release; check the LLM Deprecation and Retirement Tracker before you build on a new id.
• Rows come only from official pages. Where a provider does not publish a dated release log, its models are not listed here rather than dated from news reports.

Status key

Dates are the date on the provider's release note or announcement. Where the Google Cloud model versions page gives a release date that the Gemini API changelog does not, the row says "Vertex". Image, audio and embedding models appear only where they share a log entry with a text model family or are a major family launch.

2026 model releases by month

November 2026

As of 2026-11-15. Newest first; providers alphabetical within a date.

October 2026

As of 2026-11-15. Newest first; providers alphabetical within a date.

September 2026

As of 2026-11-15. Newest first; providers alphabetical within a date.

July and August 2026

As of 2026-11-15. Newest first; providers alphabetical within a date.

January to June 2026

As of 2026-11-15. Newest first; providers alphabetical within a date.

Coverage notes

• Not dated on the pages checked: Moonshot AI lists kimi-k3.1-pro and kimi-k3 on its pricing page without release dates, and Meta's developer models page lists muse-spark-1.3, muse-ember-1.0 and other ids without dates. These models are covered on the Moonshot AI Kimi Models and Meta Models pages.
• Not yet in this table: Alibaba Qwen, Amazon, Baidu, ByteDance, Cohere, IBM, MiniMax, NVIDIA, Tencent and Zhipu Z.ai releases are not yet listed because no dated official entry was confirmed; providers not yet listed will be added at the next update. See the Zhipu Z.ai GLM Models and Amazon Nova Models hubs in the meantime.
• Knowledge cutoff is not release date. OpenAI lists a knowledge cutoff of 30 April 2026 for GPT-6 Astra and GPT-6.1 Sol and 1 June 2026 for GPT-6.1 Terra; xAI lists May 2026 for Grok 4.7. See Knowledge Cutoff Explained.

Release volume by provider

Counts of dated rows in the tables above (text, multimodal and the image or audio families listed), 1 January to 15 November 2026. A count reflects how a provider publishes its log, not model quality or importance.

See LLM Naming and Versioning Explained for how aliases, snapshots and tiers work.

What to watch next

• Google Gemini 4 rollout: Carbon and Hydrogen have Google Cloud release dates (10 and 12 November 2026) but were not yet on the Gemini API changelog when checked; watch for GA entries there.
• Mistral Large 4: public preview from 2026-10-06; the Mistral Large 4 model card lists it as generally available from 2026-11-05, with open weights on Hugging Face under the Mistral Ecosystem Open Weight License.
• Restricted tiers: Anthropic Mythos 5.2 (verified organisations), OpenAI GPT-6 Cyber and Rosalind (approved organisations), and Google's gemini-4-argon-cyber-permissive and gemini-4-carbon-cyber-permissive ids listed on Google Cloud; access terms can change between releases.
• Alias moves: DeepSeek deepseek-flash and OpenAI gpt-5.6 (routes to gpt-5.6-sol) point to whatever the provider names; an alias change is a model change for your application.
• Kimi K3 series: Moonshot AI is retiring the K2 series by the end of December 2026, so K3 dates and ids matter for anyone on K2.

How to use this tracker

1. Pin the id, not the name. Use the API id column in configuration and record the release date beside it in your model register.
2. Check access before planning. Restricted models need an approval or verification step with the provider; build that lead time into your plan.
3. Read the source row. Each row links the provider page used. Confirm context window, pricing and regions there or on the provider hub before adoption.
4. Route through change approval. Treat a new model as a change to a production dependency. The AI Model Change Approval Template covers evaluation, sign-off and rollback.
5. Pair with retirement dates. Every adoption should record the earliest retirement date the provider states. See the LLM Deprecation and Retirement Tracker and the LLM Pricing Tracker.

Practitioner checklist

• Subscribe to each provider's release notes or changelog you depend on (the Source column lists them).
• Keep a model register with id, provider, surface (direct API or cloud), release date and access status.
• Re-run your evaluation suite when an alias you use moves to a new snapshot.
• Confirm restricted-access terms and approvals before committing a roadmap to a new model.
• Record the provider's stated retirement or "not sooner than" date at adoption time.
• Check that API keys and service accounts used for new models follow your existing controls; see the LLM API Key Security Guide.
• Confirm each row against the provider source before adoption; this tracker is updated from time to time, not on a fixed schedule.

Standards and references

• Anthropic: Claude Platform release notes
• DeepSeek: API docs change log
• Google: Gemini API release notes
• Google Cloud: Model versions and lifecycle
• Meta: Introducing Muse Wonder
• Mistral AI: Changelog
• OpenAI: API changelog
• xAI: Release notes

Related NHI Mgmt Group resources: Anthropic Models · Google Gemini Models · OpenAI Models · Foundation Model Selection Guide · AI Model Retirement Guide