Open LoopNEWS

Models

The latest models stories in AI, fact-checked and explained in plain English.

  1. Microsoft's Ignite 2026 graphic.

    Microsoft launches MAI Code 2 and six more of its own AI models at Ignite

    Microsoft AI released seven new models on 17 November 2026, led by MAI Code 2. Here is what they claim, what they cost and where you can use them.

    Confirmed · Models18 Nov 2026
  2. OpenAI's graphic for GPT-6.2 Astra and Sol.

    OpenAI releases GPT-6.2 Astra and Sol, with lower prices and better office work

    OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on 17 November 2026. Here is what they cost, how they score, who can use them and the catch.

    Confirmed · Models18 Nov 2026
  3. The words Gemini 4 Hydrogen beside a glassy blue shape on a blue background

    Google launches Gemini 4 Hydrogen and brings it to Search and the Gemini app

    Google launched Gemini 4 Hydrogen on 12 November 2026. It is in Google Search for all users and the Gemini app. API price: $1.50 per million input tokens.

    Confirmed · Models13 Nov 2026
  4. A glowing crystal beside the words Gemini 4 Carbon on a blue background

    Google makes Gemini 4 Carbon, its strongest AI model, available to developers

    Google made Gemini 4 Carbon available on 10 November 2026 to paying developers, businesses and Ultra subscribers. Price: $2 per million input tokens at launch.

    Confirmed · Models11 Nov 2026
  5. Meta's graphic reading Muse Ember is here.

    Meta launches Muse Ember, its smallest model, priced at ten cents per million input tokens

    On 9 November 2026 Meta launched Muse Ember, its smallest model, for repeat agent work. It costs $0.10 per million input tokens and reads text, images, video.

    Confirmed · Models10 Nov 2026
  6. Anthropic's graphic for Claude Fable 5.2 and Claude Mythos 5.2.

    Anthropic releases Claude Fable 5.2 and Mythos 5.2, faster and cheaper than 5.1

    On 9 November 2026 Anthropic released Claude Fable 5.2 and Mythos 5.2: up to 30% faster than 5.1, $8 per million input tokens, and a new coding record.

    Confirmed · Models10 Nov 2026
  7. A frame from Thinking Machines' demo of a city-building game made by Inkling 2 Small.

    Thinking Machines launches Inkling 2 Small, a 594B AI model with open weights promised

    Thinking Machines Lab introduced Inkling 2 Small on 5 November 2026: a 594B-parameter model, closed for now, with Apache 2.0 open weights promised next week.

    Confirmed · Models7 Nov 2026
  8. OpenAI's cover graphic for GPT-6 Cyber.

    OpenAI's GPT-6 Cyber finds software flaws, and only approved defenders can use it

    OpenAI announced GPT-6 Cyber on 5 November 2026: a security model that scores 48.5% on ExploitGym. Access starts with approved Daybreak Red customers.

    Confirmed · Models6 Nov 2026
  9. Mistral announcement banner with the Mistral logo

    Mistral releases the weights of Large 4, a 1 trillion parameter model

    Mistral released the weights of Mistral Large 4 on 29 October 2026. The 1 trillion parameter model is free to download, with limits for cloud services.

    Confirmed · Models29 Oct 2026
  10. A glowing crystal on a blue background marked with the number 4

    Google announces Gemini 4 Carbon, its strongest model, first for cyber defenders

    Google announced Gemini 4 Carbon on 28 October 2026. It calls it its strongest model. For now only a small group of cyber defenders can use it.

    Confirmed · Models29 Oct 2026
  11. A hummingbird feeding at an orange flower inside a glass bubble, next to the title Genie 4

    Google DeepMind announces Genie 4, a world model with built-in sound

    Google DeepMind announced Genie 4 on 27 October 2026: a real-time world model with native audio. It is a research preview for trusted testers.

    Confirmed · Models28 Oct 2026
  12. The GPT-6.1 Terra launch art from OpenAI, with the words GPT-6.1 and Terra beside the Earth

    OpenAI launches GPT-6.1 Terra, with API prices 50% lower than GPT-5.6 Terra

    OpenAI launched GPT-6.1 Terra on 26 October 2026. API prices are 50% lower than GPT-5.6 Terra, and it is in ChatGPT Work and Codex for paid plans.

    Confirmed · Models27 Oct 2026
  13. Dessa's two-track waveform view of a conversation with a person

    Block unveils Dessa, a voice AI trained only on human speech that runs on a laptop

    Block unveiled Dessa on 24 October 2026, a speech-to-speech AI it says learned from human conversation alone and runs locally on a MacBook Pro M4 Max.

    Confirmed · Models25 Oct 2026
  14. Kimi K3.1 Pro title card over mountains

    Moonshot AI launches Kimi K3.1 Pro, an open-weight model with weights due by 4 November

    Moonshot AI launched Kimi K3.1 Pro on 24 October 2026. It scores 65.7% on Terminal-Bench 4.0, by Moonshot's own figures. Full weights are due by 4 November.

    Confirmed · Models24 Oct 2026
  15. Google Cloud header: OpenAI models for Google Cloud customers

    OpenAI's GPT-6 models are coming to Google Cloud

    Google Cloud is adding OpenAI's GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna to its Model Garden in private preview. What it means and who can use it.

    Confirmed · Models21 Oct 2026
  16. DeepSeek's chart of prefill throughput against latency

    DeepSeek open-sources DeepInfer, the engine that runs its AI service

    DeepSeek published DeepInfer on GitHub on 20 October 2026, the engine it says runs its services. What is in it, what is missing, and the licence.

    Confirmed · Models21 Oct 2026
  17. MiniMax's chart of Automation-Bench score against cost per task

    MiniMax says its new M3.1 Flash AI costs 70% less for agent tasks

    MiniMax released M3.1 Flash on 19 October 2026. It says agent tasks cost 70% less and run nearly three times faster. Price, scores and where to get it.

    Confirmed · Models20 Oct 2026
  18. The DeepSeek logo on a dark background with streaks of light

    DeepSeek upgrades its Flash model with V4.1-Flash-1017

    On 17 October 2026 DeepSeek released V4.1-Flash-1017, a minor update with higher scores and shorter answers. The API price is unchanged.

    Confirmed · Models18 Oct 2026
  19. Muse Spark Open Weights title card with blue flowing lines

    Meta releases Muse Spark Open, an open-weight AI model that runs on two Macs

    On 15 October 2026 Meta released the weights of Muse Spark Open, an 816B-parameter model. What it is, how to get it, the licence and the Mac Studio setup.

    Confirmed · Models16 Oct 2026
  20. A large number 4 on a blue gradient background

    Google makes Gemini 4 Argon available to everyone, from $2 per million tokens

    Google made Gemini 4 Argon globally available on 13 October 2026, with API pricing from $2 per million input tokens. Here is who gets it and what is claimed.

    Confirmed · Models14 Oct 2026
  21. Meta's title card reading Introducing Muse Wonder, over rows of the word WONDER

    Meta launches Muse Wonder and Muse Wonder Full, its most capable AI models yet

    Meta introduced Muse Wonder and Muse Wonder Full on 12 October 2026, models built for agent tasks. Here is what they do, who gets them and what is next.

    Confirmed · Models13 Oct 2026
  22. Claude Haiku title card

    Anthropic releases Claude Haiku 5.5, its cheapest and fastest Claude model

    Anthropic released Claude Haiku 5.5 on 7 October 2026. It costs $0.10 per million input tokens and is available on Claude.ai, including the Free plan.

    Confirmed · Models8 Oct 2026
  23. GPT-6 with Intelligent UI title beside an interactive bike diagram

    OpenAI rolls out GPT-6 with Intelligent UI to ChatGPT users

    OpenAI began rolling out GPT-6 with Intelligent UI in ChatGPT on 7 October 2026. Paid plans get it today, Free and Go users tomorrow. Here is what changes.

    Confirmed · Models8 Oct 2026
  24. Mistral announcement banner with the Mistral logo

    Mistral launches Large 4, a 1 trillion parameter model, in preview

    Mistral launched a public preview of Mistral Large 4 on 6 October 2026, a 1 trillion parameter model, with weights promised by the end of October.

    Confirmed · Models6 Oct 2026
  25. Gemini 4 Argon key art: the words Gemini 4 Argon beside a large glowing number 4 on a blue background

    Google unveils Gemini 4 Argon, its new frontier AI model

    Google DeepMind's Gemini 4 Argon is built for long, complex work and can write 1M tokens in one answer. What it does, who gets it and what it costs.

    Confirmed · Models1 Oct 2026