AI Model Release Timeline 2026

By Versus Tools AI Team · Published · Updated · 9 min read · AI Models & News

We may earn a commission if you buy through links on this page. Read the affiliate disclosure · How we review tools

Quick answer. A 2026 AI model timeline in reverse date order. Every date below was checked on the vendor's own announcement page.

Last updated: Nov 17, 2026.

This is a reverse-chronological list of AI model launches whose dates we could read on an official page. Qwen4-Max came out Nov 16, 2026. Gemini 4 Hydrogen and Seedance 2.5 Flash came out Nov 12, 2026. Everything older in the table is here because the vendor’s own post or release index prints the date, not because a third-party calendar said so. The Nov 17 update also adds Claude Meetings, which is a product row rather than a new foundation-model id, Seedream 5.5, and Seedance 2.5 Flash.

The Seedance date is the Nov 12, 2026 announcement. BytePlus posts that day, as reported the same day, and IT Home’s Nov 12 report both say the API opens Nov 15. The model id is dreamina-seedance-2-5-flash-261115. The model-detail page we opened does not print a calendar date of its own. The pricing page that lists the id was updated Nov 16, 2026.

The page is based on official documentation, benchmarks and public reports. We did not test these models. Full write-ups: Qwen4-Max, Seedance 2.5 Flash, Gemini 4 Hydrogen, Gemini 4 Carbon, Claude Fable 5.2, Claude Meetings, and OpenAI Images 2.6. Seedream 5.5 is on GPT Image 2.6 vs Seedream 5.5. Seedance 2.5 Flash versus Gemini Omni 1.2 Flash is on Seedance 2.5 Flash vs Gemini Omni 1.2 Flash. Fable versus Terra is on Claude Fable 5.2 vs GPT-6.1 Terra. A job-by-job reading of the same sources is on best AI models.

How a row gets in

We opened the vendor page and used the date printed on it. If the announcement body says “today” and a vendor index dates that post, the index date is the one in the table, and the note says so. If we could not find a date on an official page, the model is not listed. Prices and benchmark scores are intentionally not in this table. They go stale on a different schedule and are written up in the articles.

Release table

DateCompanyModelWhat it isOfficial announcementOur article
Nov 16, 2026AlibabaQwen4-MaxFlagship mixture-of-experts model. 2.4 trillion total parameters, 71 billion active per token, native vision, 1 million-token context. API id qwen4-max.AnnouncementQwen4-Max
Nov 12, 2026GoogleGemini 4 HydrogenFast Gemini 4 model for coding and agentic workflows, beside Carbon and Argon. API id gemini-4-hydrogen. The post was updated Nov 13, 2026.AnnouncementGemini 4 Hydrogen
Nov 12, 2026ByteDanceSeedance 2.5 FlashVideo model. 480p and 720p, 4 to 30 seconds, generated audio. API id dreamina-seedance-2-5-flash-261115. The Nov 12 date is the announcement. The API was set for Nov 15. BytePlus’s pricing page, updated Nov 16, lists $7.0 per million tokens at 480p and 720p.Model pageSeedance 2.5 Flash
Nov 11, 2026BaiduERNIE 5.5 Flash PreviewFirst post-trained model on ERNIE 5.5 Flash Base. 280 billion total parameters, 14 billion active, 1 million-token context. API on Qianfan. English note says weights follow under Apache 2.0 soon. The Chinese note says by Nov 25, 2026.English release note—
Nov 10, 2026GoogleGemini 4 CarbonGeneral availability of Google’s frontier Gemini 4 model for paying developers and Google AI Ultra subscribers. A Fairwind cyber release is described as about two weeks earlier.Now availableGemini 4 Carbon
Nov 10, 2026AnthropicClaude MeetingsBeta product, not a new foundation-model id. Claude joins Google Meet as a visible participant on Team and Enterprise. Same day: Claude Whiteboards and Claude Sheets, beta on paid plans. TechCrunch reported Claude Data (formerly Claude Dashboards) became generally available.AnnouncementClaude Meetings
Nov 9, 2026AnthropicClaude Fable 5.2 and Mythos 5.2Same underlying model. Fable is the general release. Mythos is for organizations in Anthropic’s verification programs. $8 / $40 per million input and output tokens. 1 million-token context.AnnouncementClaude Fable 5.2
Nov 9, 2026MetaMuse EmberMeta’s smallest model in this lineup, for high-volume agent work. API id muse-ember-1.0. $0.10 / $0.50 per million input and output tokens. 1 million-token context, up to 512K output tokens.Announcement—
Nov 9, 2026ByteDanceSeedream 5.5Image model. The Seed blog describes more stable multi-turn editing and native RGBA output. The head-to-head with GPT-Image-2.6 has its own page.Seed blogGPT Image 2.6 vs Seedream 5.5
Nov 5, 2026GoogleGemini Omni 1.2 FlashVideo model. Model id gemini-omni-1.2-flash. GA Nov 5, 2026. Native clips up to 15 seconds. Listed price $0.07 per second for 720p, which Google says is 30 percent below Omni 1.1. Veo 3.1 is the older Google video model.AnnouncementSeedance 2.5 Flash vs Gemini Omni 1.2 Flash
Nov 5, 2026OpenAIGPT-6 CyberCybersecurity model for authorized vulnerability research. Initially available through the API and Codex to approved Daybreak Red customers. The launch post does not list a token price.Announcement—
Nov 5, 2026BaiduERNIE 5.5 Flash BasePre-training checkpoint, before mid-training or post-training. 280 billion total parameters, 14 billion active. Weights posted for further training.Release note—
Nov 4, 2026TencentHy4Tencent Hunyuan model with vision and computer use. Tencent describes it as open-weight. API price on the research post: $0.834 input, $2.501 output, and $0.042 cached input, per million tokens.Research post—
Nov 4, 2026OpenAIImages 2.6Image model in ChatGPT, Codex, and the API, including GIF generation. API model names in the post: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. Token price is not printed on the announcement.AnnouncementOpenAI Images 2.6
Oct 26, 2026OpenAIGPT-6.1 TerraFast model for computer use, visual reasoning, and writing. API id gpt-6.1-terra. Standard price $1 input, $0.05 cached input, and $5 output per million tokens. The date is on OpenAI’s research release index and API changelog.AnnouncementClaude Fable 5.2 vs GPT-6.1 Terra
Oct 24, 2026MoonshotKimi K3.1 ProOpen-weight model Moonshot calls 3T-class, with a 1 million-token context. Blog index date is 2026-10-24. The post said full weights would be released by Nov 4, 2026. The public model card lists 2.8 trillion total parameters and 104 billion active.Announcement—

Same-day rows follow the order we verified them, with language-model launches before the image launch on Nov 4.

What we left out

GPT-6.1 Terra’s announcement page, as fetched, does not print a date on the headline. The Oct 26, 2026 date is taken from two other OpenAI pages: the research release index and the API changelog entry for gpt-6.1-terra. Kimi’s date is the blog index, not a line inside the post.

We did not add a row for models that showed up only on unofficial timelines. DeepSeek-V4.1-Flash-1017 is the version name on DeepSeek’s current pricing page, and that page does not print a calendar launch date, so it is not in the table. Its price is discussed on the Qwen4-Max comparison because the pricing page itself is official.

Mistral Large 4 was announced Oct 6, 2026. Our earlier article is Mistral Large 4. It is not restated in the seed table above.

The rows above are models. These are tools that sit in front of a model:

Sources

Each row links to the page that carries its date. Supporting pages used for a price or a model id in the “what it is” column:

Last updated: Nov 17, 2026. This page contains affiliate links.

Frequently asked questions

Which AI models launched in November 2026?

On official pages checked Nov 16, 2026, with rows added or linked on Nov 17, 2026, the November entries are Qwen4-Max on Nov 16, Gemini 4 Hydrogen and Seedance 2.5 Flash on Nov 12, ERNIE 5.5 Flash Preview on Nov 11, Gemini 4 Carbon and Claude Meetings on Nov 10, Claude Fable 5.2 and Mythos 5.2, Muse Ember, and Seedream 5.5 on Nov 9, Gemini Omni 1.2 Flash, GPT-6 Cyber, and ERNIE 5.5 Flash Base on Nov 5, and Hy4 and Images 2.6 on Nov 4. Claude Meetings is a product row, not a new foundation-model id. The Seedance 2.5 Flash date is the Nov 12 announcement. The API id is dreamina-seedance-2-5-flash-261115.

Why is Gemini 4 Carbon dated Nov 10, 2026?

Nov 10, 2026 is the date on Google's post titled Gemini 4 Carbon is now available, which opened the model to paying developers and to Google AI Ultra subscribers. That post also says a narrower cyber release to defenders in the Fairwind Program happened about two weeks earlier. This timeline uses the dated public-availability post, and it does not add a separate preview row without its own dated page.

Why are some 2026 models missing?

A row is included only when we could read a date on an official page. Secondary launch calendars were not used. Older 2026 models are not backfilled here yet. Kimi K3.1 Pro is included because Moonshot's blog index dates the post 2026-10-24, even though the post body says today.

Did you test the models on this list?

No. The table is a date tracker based on official documentation, benchmarks and public reports. Write-ups of what a model costs or how it scores live on the linked articles, and only where the vendor published the number.

Where can I read more than the date?

We have full articles for Qwen4-Max (Nov 16, 2026), Seedance 2.5 Flash (announced Nov 12, 2026), Gemini 4 Hydrogen (Nov 12, 2026), Claude Meetings (Nov 10, 2026), Gemini 4 Carbon (Nov 10, 2026), Claude Fable 5.2 and Mythos 5.2 (Nov 9, 2026), and OpenAI Images 2.6 (Nov 4, 2026). Seedream 5.5 has a head-to-head page against GPT-Image-2.6. Seedance 2.5 Flash has a head-to-head page against Gemini Omni 1.2 Flash, Google's Nov 5, 2026 video model. Claude Fable 5.2 has a head-to-head page against GPT-6.1 Terra. The November ranking page sorts the current field by job. Mistral Large 4, announced Oct 6, 2026, has its own earlier article and is not repeated in this seed table.

When was this timeline last updated?

Last updated: Nov 17, 2026. That update added Claude Meetings, Seedream 5.5, and Seedance 2.5 Flash, linked the OpenAI Images 2.6 article, pointed the Seedream row at the GPT Image 2.6 vs Seedream 5.5 comparison, linked Claude Fable 5.2 vs GPT-6.1 Terra, and pointed the Gemini Omni 1.2 Flash row at the Seedance comparison. Qwen4-Max's announcement date is still Nov 16, 2026. The older rows were read on Nov 16, 2026.