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
Google · Gemini API
The last available tuning model, Gemini 1.5 Flash 001, has been shut down. Tuning is no longer supported on…
The last available tuning model, Gemini 1.5 Flash 001, has been shut down. Tuning is no longer supported on any models.
Mistral AI · Mistral API
We released the new Agents API
We released the new Agents API.
Anthropic · Claude API
We’ve launched Claude Opus 4 and Claude Sonnet 4, our latest models with extended thinking capabilities
Learn more in Models overview. The default behavior of extended thinking in Claude 4 models returns a summary of Claude’s full thinking process, with the full thinking encrypted and returned in the signature field of thinking block output. We’ve launched interleaved thinking in beta, a feature that enables Claude to think in between tool calls. We’ve launched the Files API in beta, enabling you to upload files and reference them in the Messages API and code execution tool. We’ve launched the Code execution tool in beta, a tool that enables Claude to execute Python code in a secure, sandboxed environment. We’ve launched the MCP connector in beta, a feature that allows you to connect to remote MCP servers directly from the Messages API. To increase answer quality and decrease tool errors, we’ve changed the default value for the top_p nucleus sampling parameter in the Messages API from 0.999 to 0.99 for all models. Our Go SDK has moved from beta to its first stable release. We’ve added minute and hour level granularity to the Usage page of Console alongside 429 error rates on the Usage page.
OpenAI · Codex app
Reworked environment page
It's now easier and faster to set up code execution. . Fixes & improvements. Added a button to retry failed tasks. Added indicators to show that the agent runs without network access after setup. Added options to copy git patches after pushing a PR. Added support for unicode branch names. Fixed a bug where secrets were not piped to the setup script. Fixed creating branches when there’s a branch name conflict. Fixed rendering diffs with multi-character emojis. Improved error messages when starting tasks, running setup scripts, pushing PRs, or disconnected from GitHub to be more specific and indicate how to resolve the error. Improved onboarding for teams. Polished how new tasks look while loading. Polished the followup composer. Reduced GitHub disconnects by 90%. Reduced PR creation latency by 35%. Reduced tool call latency by 50%. Reduced task completion latency by 20%. Started setting page titles to task names so Codex tabs are easier to tell apart. Tweaked the system prompt so that agent knows it’s working without network, and can suggest that the user set up dependencies. Updated the docs.
Mistral AI · New model
We released Mistral OCR 2 and annotations
We released Mistral OCR 2 (mistral-ocr-2505) and annotations.
Anthropic · Claude API
Our Ruby SDK has moved from beta to its first stable release
Mistral AI · New model
We released Devstral Small
We released Devstral Small (devstral-small-2505).
Google · Gemini API
Launched support for custom video preprocessing using clipping intervals and configurable frame rate sampling
Launched support for custom video preprocessing using clipping intervals and configurable frame rate sampling. Launched multi-tool use, which supports configuring code execution and Grounding with Google Search on the same generateContent request. Launched support for asynchronous function calls in the Live API. Launched an experimental URL context tool for providing URLs as additional context to prompts. Released gemini-2.5-flash-preview-05-20, a Gemini preview model optimized for price-performance and adaptive thinking. To learn more, see Gemini 2.5 Flash Preview and Thinking. Released the gemini-2.5-pro-preview-tts and gemini-2.5-flash-preview-tts models, which are capable of generating speech with one or two speakers. Released the lyria-realtime-exp model, which generates music in real time. Released gemini-2.5-flash-preview-native-audio-dialog and gemini-2.5-flash-exp-native-audio-thinking-dialog, new Gemini models for the Live API with native audio output capabilities. To learn more, see the Live API guide and Gemini 2.5 Flash Native Audio. Released gemma-3n-e4b-it preview, available on AI Studio and through the Gemini API, as part of the Gemma 3n launch.
Google · Gemini app
Create custom practice quizzes with Gemini, plus 8 more changes
What: Starting today, you can generate custom practice quizzes to help you prepare for an upcoming exam or simply increase your knowledge of any topic - big or small. Create quizzes based on documents you want to study, such as PDFs or class notes, or ask Gemini to create a quiz on a specific topic, and you’ll get a dynamic quiz experience, complete with hints, explanations for right and wrong answers, and a helpful summary at the end highlighting where you did well and where you may need to study a little harder. This experience is available to users over the age of 18 and to qualifying Google Workspace education plans. Why: With Gemini’s intuitive and streamlined practice quizzes, learners can lean on the advantages of generative AI to help them prepare: unlimited quiz generation, personalized responses, and conversational learning experiences.
OpenAI · Codex app
Codex in the ChatGPT iOS app
Start tasks, view diffs, and push PRs—while you're away from your desk. .
Anthropic · Claude API
We’ve launched a web search tool in the API, allowing Claude to access up-to-date information from the web
Learn more in Web search tool.
Google · Gemini API
Released gemini-2.0-flash-preview-image-generation, a preview model for generating and editing images. To…
Released gemini-2.0-flash-preview-image-generation, a preview model for generating and editing images. To learn more, see Image generation and Gemini 2.0 Flash Preview Image Generation.
Mistral AI · New model
We released Mistral Medium 3
We released Mistral Medium 3 (mistral-medium-2505).
Google · Gemini API
Released gemini-2.5-pro-preview-05-06, a new version of our most powerful model, with improvements on code…
Released gemini-2.5-pro-preview-05-06, a new version of our most powerful model, with improvements on code and function calling. gemini-2.5-pro-preview-03-25 will automatically point to the new version of the model.
Google · Gemini app
2.5 Pro got a major upgrade for coding
What: We’re rolling out an updated version of Gemini 2.5 Pro with improved capabilities for coding. This model more intuitively understands coding prompts and produces stronger outputs, enabling users to quickly create compelling web apps in Canvas. This improvement builds on the overwhelmingly positive feedback to 2.5 Pro’s coding and multimodal reasoning capabilities. Why: We believe in rapid iteration and bringing the best of Gemini to the world. Your feedback helps us improve these models over time and learning from experimental launches informs how we release models more widely.
Anthropic · Claude API
Cache control must now be specified directly in the parent content block of tool_result and document.source
For backwards compatibility, if cache control is detected on the last block in tool_result.content or document.source.content, it will be automatically applied to the parent block instead. Cache control on any other blocks within tool_result.content and document.source.content will result in a validation error.
xAI · Grok API
Cached prompt is now available
You can now use cached prompt to save on repeated prompts. For more info, see models.
xAI · Grok API
Live Search is available on API
Live search is now available on API. Users can generate completions with queries on supported data sources.
Google · Gemini app
Create and share videos with Veo 2 in Gemini Advanced
What: Starting today, Gemini Advanced subscribers can generate high-quality, 8-second videos using Veo 2. Simply describe what you have in mind and watch your ideas come to life in motion - whether you're creating for fun, sharing with friends, or adding a dynamic element to your projects. How to access: Select ‘Veo 2’ from the model dropdown. How to share: Download the MP4 file, generate a public link, or share the video directly (mobile only).
Google · Gemini app
Update to 2.0 Flash in Gemini
What: An enhanced version of Gemini 2.0 Flash is now available to Gemini app users. This update enables a more natural, collaborative, and adaptive conversational style, providing users with more fluid and engaging interactions that feel in tune with specific topics and tasks. Try chatting about your interests; working through a problem at school, work, or home; or asking for a new, more creative perspective to feel the difference. Why: We’re committed to making Gemini a more insightful, collaborative partner. When it’s better at understanding context and engaging in helpful dialogue, it’s easier for you to create, interact, and accomplish your goals.
Google · Gemini API
Released gemini-2.5-flash-preview-04-17, a Gemini preview model optimized for price-performance and adaptive…
Released gemini-2.5-flash-preview-04-17, a Gemini preview model optimized for price-performance and adaptive thinking. To learn more, see Gemini 2.5 Flash Preview and Thinking.
Google · Gemini app
Try our 2.5 Flash model
What: Starting today, Gemini app users have access to our 2.5 Flash (experimental) model - a fast and efficient thinking model that also provides strong performance. It’s ideal for tasks that require advanced reasoning and benefit from faster processing, like summarization, document analysis and data extraction. This experimental model is meant to be an early preview and can have unexpected behaviors and may make mistakes. Why: We believe in rapid iteration and bringing the best of Gemini to the world. Your feedback helps us improve these models over time and learning from experimental launches informs how we release models more widely.
Google · Gemini API
Launched context caching for Gemini 2.0 Flash
Launched context caching for Gemini 2.0 Flash.
Mistral AI · Mistral API
We released the Classifier Factory
We released the Classifier Factory.
Anthropic · Claude API
We launched a beta version of the Ruby SDK
Google · Gemini API
Released veo-2.0-generate-001, a generally available text- and image-to-video model, capable of generating…
Released veo-2.0-generate-001, a generally available (GA) text- and image-to-video model, capable of generating detailed and artistically nuanced videos. To learn more, see the Veo docs. Released gemini-2.0-flash-live-001, a public preview version of the Live API model with billing enabled. Enhanced Session Management and Reliability Session Resumption: Keep sessions alive across temporary network disruptions. The API now supports server-side session state storage (for up to 24 hours) and provides handles (session_resumption) to reconnect and resume where you left off. Longer Sessions via Context Compression: Enable extended interactions beyond previous time limits. Configure context window compression with a sliding window mechanism to automatically manage context length, preventing abrupt terminations due to context limits. Graceful Disconnect Notification: Receive a GoAway server message indicating when a connection is about to close, allowing for graceful handling before termination. More Control over Interaction Dynamics Configurable Voice Activity Detection (VAD): Choose sensitivity levels or disable automatic VAD entirely and use new client events (activityStart, activityEnd) for manual turn control. Configurable Interruption Handling: Decide whether user input should interrupt the model's response. Configurable Turn Coverage: Choose whether the API processes all audio and video input continuously or only captures it when the end-user is detected speaking. Configurable Media Resolution: Optimize for quality or token usage by selecting the resolution for input media. Richer Output and Features Expanded Voice & Language Options: Choose from two new voices and 30 new languages for audio output. The output language is now configurable within speechConfig. Text Streaming: Receive text responses incrementally as they are generated, enabling faster display to the user. Token Usage Reporting: Gain insights into usage with detailed token counts provided in the usageMetadata field of server messages, broken down by modality and prompt or response phases.
Google · Gemini API
Released gemini-2.5-pro-preview-03-25, a public preview Gemini 2.5 Pro version with billing enabled. You can…
Released gemini-2.5-pro-preview-03-25, a public preview Gemini 2.5 Pro version with billing enabled. You can continue to use gemini-2.5-pro-exp-03-25 on the free tier.
xAI · New model
Grok 3 models launch on API
Our latest flagship Grok 3 models are now generally available via the API. For more info, see models.
Anthropic · Claude API
Our Java SDK has moved from beta to its first stable release
We’ve moved our Go SDK from alpha to beta.
Google · Gemini app
Our 2.5 Pro model is now available to all Gemini users, with Canvas
What: Today we’re expanding access to our most intelligent AI model, 2.5 Pro (experimental), to all Gemini users. This state-of-the-art model has thinking capabilities natively built in, with exceptional performance in coding, math, image understanding and more. Canvas, our new interactive space that makes it easy to create, refine, and share your work, is now available to try with 2.5 Pro (experimental). With 2.5 Pro’s improved coding capabilities, coupled with Canvas, you can quickly create compelling web apps or generate code for immediate use. Gemini users will be able to try 2.5 Pro (experimental) with rate limits, and Gemini Advanced users will continue to have expanded access and a significantly larger context window. Being an experimental model, it can have unexpected behaviors and may make mistakes. Why: We want to bring the best model in the world to all Gemini users. Your feedback helps us improve these models over time and learning from experimental launches informs how we release models more widely.