AI posts
AI release notes
Official updates from Anthropic, OpenAI, Google, xAI, Mistral, DeepSeek and Meta, newest first. Open any update for a short AI summary and the full notes.
Compiled by the Toolstash team · Updated
Mistral AI · New model
We released Mistral Small 3
We released Mistral Small 3 (mistral-small-2501).
Mistral AI · Mistral API
We released custom structured outputs for all models
We released custom structured outputs for all models.
Anthropic · Claude API
We’ve launched citations capability in the API, allowing Claude to provide source attribution for information
Learn more in Citations. We’ve added support for plain text documents and custom content documents in the Messages API.
Anthropic · Claude API
We announced the deprecation of the Claude 2, Claude 2.1, and Claude Sonnet 3 models
Read more in Model deprecations.
Google · Gemini API
Released gemini-2.0-flash-thinking-exp-01-21, the latest preview version of the model behind the Gemini 2.0…
Released gemini-2.0-flash-thinking-exp-01-21, the latest preview version of the model behind the Gemini 2.0 Flash Thinking Model.
DeepSeek · New model
deepseek-reasoner
deepseek-reasoner is our new model DeepSeek-R1. You can invoke DeepSeek-V3 by specifying model='deepseek-reasoner'.. For details, please refer to: DeepSeek-R1 Release. For guides, please refer to: Thinking Mode
Anthropic · Claude API
We’ve updated prompt caching to be easier to use
Now, when you set a cache breakpoint, we’ll automatically read from your longest previously cached prefix. You can now put words in Claude’s mouth when using tools.
Mistral AI · New model
We released Codestral 2501
We released Codestral 2501 (codestral-2501).
Anthropic · Claude API
We’ve optimized support for prompt caching in the Message Batches API to improve cache hit rate
xAI · Grok API
Docs Dark Mode
Released dark mode support on docs.x.ai
xAI · Grok API
Status Page
Check service statuses across all SpaceXSI products at status.x.ai.
DeepSeek · New model
deepseek-chat
The deepseek-chat model has been upgraded to DeepSeek-V3. The API remains unchanged. You can invoke DeepSeek-V3 by specifying model='deepseek-chat'.. For details, please refer to: introducing DeepSeek-V3
Anthropic · Claude API
We’ve added support for a delete endpoint in the Message Batches API
Google · Gemini API
Released Gemini 2.0 Flash Thinking Mode for public preview. Thinking Mode is a test-time compute model that…
Released Gemini 2.0 Flash Thinking Mode for public preview. Thinking Mode is a test-time compute model that lets you see the model's thought process while it generates a response, and produces responses with stronger reasoning capabilities. Read more about Gemini 2.0 Flash Thinking Mode in our overview page.
Google · Gemini app
Priority access with Gemini Advanced: Upload your code repository
What: In September, we made it possible for you to upload multiple code files to get help understanding, modifying, or debugging your code within Gemini Advanced. Today, we're excited to announce yet another upgrade: you can now upload your code repository (up to 1 folder per conversation, with a maximum of 1,000 files and a maximum folder size of 100MB, can be uploaded at a time) directly from your device in one action. This means you'll get the same comprehensive code assistance, with the added benefit of Gemini Advanced understanding the full context of your project. The experience is available with Gemini Advanced as part of the Google One AI Premium Plan. Why: This improvement enables you to streamline your workflow and leverage the full power of Gemini Advanced's 1M token context window for more comprehensive and contextually relevant code analysis and debugging.
Anthropic · Claude API
Models API: Query available models, validate model IDs, and resolve model aliases to their canonical model IDs
Message Batches API: Process large batches of messages asynchronously at 50% of the standard API cost. Token counting API: Calculate token counts for Messages before sending them to Claude. Prompt Caching: Reduce costs by up to 90% and latency by up to 80% by caching and reusing prompt content. PDF support: Process PDFs to analyze both text and visual content within documents. Java SDK (alpha) Go SDK (alpha)
Google · Gemini app
Priority access with Gemini Advanced: Try our latest 2.0 Experimental Advanced model
What: Gemini Advanced users can now try Gemini-exp-1206, our model designed to be exceptional at complex tasks such as coding, math, reasoning & instruction following. Whether you’re tackling advanced coding challenges, solving mathematical problems for school or personal projects, or providing detailed, multi-step instructions to craft a tailored business plan, 2.0 Experimental Advanced will help you navigate complex tasks with greater ease. This experimental model is meant to be an early preview and can have unexpected behaviors and may make mistakes. Additionally this model will not have access to real time information and won't be compatible with some Gemini features in its experimental state. You can access this experimental model in the Gemini model drop-down. Why: We believe in rapid iteration and bringing the best of Gemini to the world, and we want to give Gemini Advanced subscribers priority access to our latest AI innovations. Which is why we’re introducing Gemini-exp-1206 to enable users to experiment with cutting-edge models with our latest enhancements. Your feedback helps us improve these models over time and learning from experimental launches informs how we release models more widely.
Google · Gemini API
Released Gemini 2.0 Flash Experimental for public preview. Gemini 2.0 Flash Experimental's partial list of…
Released Gemini 2.0 Flash Experimental for public preview. Gemini 2.0 Flash Experimental's partial list of features includes: Twice as fast as Gemini 1.5 Pro Bidirectional streaming with our Live API Multimodal response generation in the form of text, images, and speech Built-in tool use with multi-turn reasoning to use features like code execution, Search, function calling, and more
Google · Gemini app
Introducing Gemini 2.0 Flash as an experimental model in Gemini, plus 1 more change
What: Gemini users now have access to our latest experimental model, a chat optimized version of 2.0 Flash Experimental with improvements on a number of key academic benchmarks and speed. This experimental model is meant to be an early preview and can have unexpected behaviors and may make mistakes. Additionally, some Gemini features won't be compatible with this model in its experimental state. You can access this experimental model in the Gemini model drop-down. Why: We believe in rapid iteration and bringing the best of Gemini to the world. Which is why we want to give users access to cutting-edge models with our latest enhancements. Your feedback helps us improve these models over time and learning from experimental launches informs how we release models more widely.
DeepSeek · New model
deepseek-chat
The deepseek-chat model has been upgraded to DeepSeek-V2.5-1210, with improvements across various capabilities. Relevant benchmarking results include: Additionally, the new version of the model has optimized the user experience for file upload and webpage summarization functionalities.
Google · Gemini app
Additional premade Gems available for Google Workspace customers
What: We're excited to announce the launch of a series of new premade Gems, designed to give Gemini for Google Workspace subscribers access to AI-powered expertise across a variety of work related use cases, starting in English. The list of premade Gems includes Hiring consultant, Copy creator, Sales pitch ideator, Sentiment analyzer, and Outreach specialist. Why: Business users need an AI solution with a deeper understanding of core business outcomes, and a focus on formulas, metrics and role profiles. These Gems provide advanced capabilities to help you tackle complex tasks and streamline your workflow.
Google · Gemini app
Priority access with Gemini Advanced: Play chess with our new Chess Gem
What: The Chess champ Gem brings a new, conversational way to experience chess within the Gemini app. Explore different openings, banter with your AI opponent, and see if you can beat it in a game. This experimental Gem is available in English in the Gemini web app for Gemini Advanced subscribers via Google One AI Premium Plan. Why: Experience chess in a whole new way! Chess champ lets you test your talent against a language model. Enjoy a full game of chess and chatting.
Anthropic · Claude API
We’ve added the ability to group by API key on the Usage and Cost pages of the Developer Console
We’ve added two new Last used at and Cost columns and the ability to sort by any column on the API keys page of the Developer Console.
Google · Gemini app
Expanding Workspace extensions to more languages: Keep and Tasks
What: Over the coming days, more users around the world will be able to connect with more apps and services in 9 languages (Korean, Japanese, Spanish, French, German, Bahasa Indonesia, Hindi, Portuguese, and Chinese), letting you set reminders with Google Tasks, make and refine lists with Google Keep, and engage with playlists in YouTube Music. Learn more about extensions. Why: We're continuing to build new ways for Gemini to connect to the apps you rely on every day. Through these extensions, Gemini can coordinate across your apps and take actions for you, better than ever. Read more about how Gemini works with your favorite Assistant actions, including details on upcoming improvements.
xAI · Grok API
Replit & SpaceXAI
Replit Agents can now integrate with SpaceXAI! Start empowering your agents with Grok. Check out the announcement for more information.
xAI · Grok API
Replit & SpaceXSI
Replit Agents can now integrate with SpaceXSI! Start empowering your agents with Grok. Check out the announcement for more information.
xAI · Grok API
Replit & xAI
Replit Agents can now integrate with xAI! Start empowering your agents with Grok. Check out the announcement for more information.
xAI · Grok API
Structured Outputs
We're excited to announce that Grok now supports structured outputs. Grok can now format responses in a predefined, organized format rather than free-form text. 1.
xAI · Grok API
Tokenizer Playground
Understanding tokens can be hard. Check out console.x.ai to get a better understanding of what counts as a token.
xAI · New model
Released the new grok-2-1212 and grok-2-vision-1212 models
A month ago, we launched the public beta of our enterprise API with grok-beta and grok-vision-beta. We’re adding grok-2-1212 and grok-2-vision-1212, offering better accuracy, instruction-following, and multilingual capabilities.