AIToolStack releases
- 14posts a week
- 4 Novlast post
Dated public releases for tools on the AIToolStack list, linked to the official announcement.
Subscribe
Latest posts
Tencent Hy4
Tencent announced Hy4, an open-source model with visual understanding and computer use. The announcement says the API is approximately $0.90 per million input tokens and $2.70 per million output tokens, and that weights and system cards are planned in the coming days. Performance comparisons on…
Meituan LongCat-2.5
Meituan released LongCat-2.5, a native multimodal model for long-horizon agent tasks, with 1.6 trillion total parameters. The company says people can try it at longcat.ai and developers can call it through an API. Vendor source: https://www.meituan.com/news/NN261102280001343
DiffuSpace Clover-35B-A3B-Preview
DiffuSpace released Clover-35B-A3B-Preview, an open-source Apache-licensed mixture-of-experts diffusion language model with 35 billion total parameters and 3 billion active, aimed at agentic coding and tool use. Benchmark and speed figures on the post are vendor-reported. Vendor source…
TBC mischief 1
The Biological Computing Co. launched TBC mischief 1, an image-to-video model built on Wan 2.2 and available on fal and OpenRouter. The company says it is 5 times faster than its base model at half a cent per second; those speed and price figures are vendor claims. Vendor source…
Mistral Large 4 open weights
Mistral released open weights for Mistral Large 4 (Le Chonk), a 1-trillion-parameter multimodal mixture-of-experts model, under the Mistral Ecosystem Open-Weight License, a modified MIT license that requires a separate commercial license for Model-as-a-Service businesses. The API price on the page…
Pika H3 Flex
Pika launched Pika H3 Flex, an open-source video and audio model family post-trained on MiniMax H3, with weights on Hugging Face. Pika says it ranks number 1 among open-source text-to-video models; that ranking is a vendor claim. Vendor source: https://pika.art/blog/pika-h3-flex-research
Gemini app file editing
Google added file generation, preview, and editing in the Gemini app for documents, PDFs, slides, spreadsheets, and code. PDF preview and editing work on web and mobile; Google says the other formats are web only, and that this is available to all users where the Gemini app is available. Vendor…
Gemini 4 Carbon, limited access
Google announced Gemini 4 Carbon, and it is not generally available. Access starts with a subset of cyber defenders through the Fairwind Program, with wider availability later, at an introductory $2 per million input tokens and $10 per million output tokens.
Particle-1
Particle launched Particle-1, a coding model post-trained on DeepSeek V4.1 Flash and available through the Particle API. Particle reports that it uses around 45% fewer tokens on average; that figure is vendor-reported.
GitHub Copilot Fusion
is generally available in the Copilot model picker, with Standard, High, and Max effort and Single, Cascade, and Critique patterns, on all Copilot plans in VS Code, the Copilot app, Copilot CLI, and JetBrains. GitHub says usage is billed per model used, with a 30% discount for…
NVIDIA Nemotron 3.5 Transcribe
NVIDIA published Nemotron 3.5 Transcribe, an open-weight promptable speech model (30B-A4B, about 4.3 billion active parameters) for transcription, speaker attribution, translation, and timestamps, under the OpenMDW-1.1 license. Leaderboard and speed figures on the page are vendor-reported. Vendor…
iCloud Mail and Calendar in ChatGPT
OpenAI's English release notes for October 27, 2026 say iCloud users can connect Mail and Calendar in ChatGPT to find emails, catch up on an inbox, check a schedule, and update calendar events. It is available on all ChatGPT plans on the web, the desktop app, and iOS.
Genie 4 research preview
Google DeepMind announced Genie 4, a world-simulation model, as a research preview for trusted testers. A waitlist is open for Google AI Ultra users in Project Genie before a later release to consumers and developers.
Liquid LFM3
Liquid AI announced LFM3, a family of open-weight models in the 2B parameter range covering text, vision, audio, omni, and decision models. The weights are on Hugging Face, and the models are also on the Liquid API.
Ling-3.1-flash-1019
inclusionAI published Ling-3.1-flash-1019 under the MIT license, with 560 billion total parameters, 25 billion active per token, and a 1 million token context window. Benchmark gains on the card are vendor-reported, and this date is the Hugging Face repository creation time because the card does…
GPT-6.1 Terra
OpenAI introduced GPT-6.1 Terra for ChatGPT Work and Codex, and in the API as gpt-6.1-terra at $1 per million input tokens and $5 per million output tokens. The English release notes date the rollout October 26, 2026, include Plus, Pro, Business, Enterprise, and Edu, and say Terra is available in…
Nanbeige4.3-33B
Nanbeige LLM Lab published Nanbeige4.3-33B, open weights under the Apache 2.0 license with a context of up to 512K tokens. Benchmark scores on the card are vendor-reported. The card does not print a calendar date; this date is the Hugging Face repository creation time.
NVIDIA SkillSpector v2.13.1
NVIDIA published SkillSpector v2.13.1, an open-source security scanner for AI agent skills under the Apache 2.0 license.
Kimi K3.1 Pro and Highspeed on Vercel AI Gateway
Kimi K3.1 Pro and Kimi K3.1 Pro Highspeed are available on Vercel AI Gateway. Vercel describes Highspeed as a faster version for interactive use; K3.1 Pro keeps a 1M-token context window and image input, and the changelog says Moonshot plans to release the weights by November 4.
Canva Craft
Canva released Craft, its first image model, in Canva and through the public API. It generates images at up to 2K resolution across more than 30 aspect ratios and can render text in 55 languages.
In the directory since 26 Oct 2026 · last checked 9 Nov 2026 · RSS