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
Anthropic · Claude API
The Messages API can now compact a conversation on demand on the Claude API, in beta with the…
Send the top-level compaction parameter, and the API returns a signed compaction block that summarizes the messages you sent. On later requests, send that block first, in place of those messages. You choose when to compact, the request can run in the background, and you can keep recent turns word for word after the summary. On models with preserved thinking, the thinking in those kept turns can stay valid. With the thinking-binding-controls-2026-08-01 beta header, the input_transformations response field gains a second entry type, thinking_mismatch_allowed.
OpenAI · Codex app
GPT-5.3-Codex-Spark deprecated
GPT-5.3-Codex-Spark was deprecated on September 14, 2026. The research preview is no longer available in the ChatGPT desktop app, Codex CLI, or IDE extension. Switch to one of the recommended models, and update saved configurations, custom agents, and scripts that select gpt-5.3-codex-spark. For faster responses with a supported model, see Fast mode.
OpenAI · Codex app
GPT-5.5 retires from ChatGPT, ChatGPT Work, and Codex on October 14
On October 14, 2026, GPT-5.5 will retire from ChatGPT, ChatGPT Work, and Codex on all plans, including consumer, Business, Enterprise, and Edu plans. This retirement does not apply to the OpenAI API. For Codex with ChatGPT sign-in, choose an available replacement for your plan and client. Update workspace defaults, saved model settings, managed configurations, custom agents, scheduled tasks, and scripts that still select gpt-5.5.
Google · Antigravity
Conversation compaction controls, forward-looking budget scopes, and tool output token truncation
This release introduces first-class conversation compaction controls via CompactionConfig, delta and forward-looking budget scopes for session resumption, tool output token truncation limits, and expanded arithmetic operations on UsageMetadata. It also delivers OS-level terminal sandboxing examples and automated tool wrapper reflection preservation.
Anthropic · Claude Code
Claude Code 2.1.270
1 fixFixed read-only git commands in Bash unexpectedly asking for permission after a session had been running for a while (regression in 2.1.269).
Google · Antigravity
Plugin MCP namespacing, conversation memory leak fixes, oversized file guardrails, and unsandboxed rule…
Enhances startup migration diagnostics for unsandboxed permission rules, speeds up /resume with stale caches, automatically namespaces plugin MCP servers to avoid collision, resolves step-cache goroutine leaks, and fixes critical issues with Gemini API thinking block retention, oversized binary file loading, artifact reviews, and database cleanup.
Anthropic · Claude Code
Claude Code 2.1.269
14 new66 fixes18 otherAdded claude plugin eval: run a plugin's eval suite against Claude Code and get scored, reproducible results (JSON + HTML report); see claude plugin eval --help. Added /output-style [name] to list and switch output styles, including over Remote Control and in cloud and other headless sessions.
OpenAI · Codex app
Quick chats with Pets and Appshots on Windows
Keep ChatGPT close while you work in other apps. In the ChatGPT desktop app on macOS and Windows, you can now type a quick chat from the floating Pets controls and press Enter to send it. Use @ to add context and $ to choose a skill. Select the bell icon to follow progress in your threads, then select a thread to open the full conversation. Choose a companion or show just the controls in Settings > Pets. Press Option+Space on macOS or Windows+Alt+P on Windows to show the controls and focus Quick Chat. Pressing the shortcut again keeps Quick Chat open. You can change Show pet in Settings > Keyboard shortcuts. To hide the controls, select Hide pet in the command menu or Settings > Pets. On macOS, you can also send an appshot to the floating controls. Keep the main ChatGPT window in the background and choose Automatic for Appshot destination. Learn more about Pets. Appshots are now available in the ChatGPT desktop app on Windows. Press both Alt keys to share the frontmost app window with ChatGPT, then ask a question or describe what you want to do. An appshot includes a screenshot and available text from that window. You can customize the shortcut and choose which chat receives it in settings. Other improvements and bug fixes. Open files directly from a conversation's Sources panel, or download files that ChatGPT can't preview. Assign a Codex Micro key to Insert text and add your own reusable text to the active prompt without sending it. Reset a pet to its default size in settings. Custom pets also stay in place when you refresh their artwork. Keep unfinished comments on response text when you switch chats and return. ChatGPT dictation now respects your saved Main language preference. Keep tab widths and scroll positions more stable as you close browser tabs in full and split views.
Google · Antigravity
Custom agent component exclusions, resilient model API retry handling, and conversation continuation fixes
Introduces excludeDefaultComponents for custom agent Markdown definitions, adds automated exponential backoff retries for transient model API errors, improves MCP open object schema validation, and fixes issues with --continue workspace resolution, inline terminal screen flashes, Remote Control auth display, and sandbox status indicators.
Anthropic · Claude Code
Claude Code 2.1.268
11 new59 fixes26 otherAdded to the Claude apps gateway: with pricing: set in gateway.yaml, signed-in Claude Code clients receive the same rates through managed settings, so /cost and telemetry match the spend meter. Added a startup warning for gateways when access_control.allow_cidrs is empty, and a one-time warning the first time a request arrives from a public address.
Anthropic · Claude API
Claude Managed Agents permission policies now include auto: the server evaluates each agent or MCP tool call…
Version 1.32.0 of the ant CLI adds ant beta:sessions connect, which attaches your terminal to a Claude Managed Agents session.
Anthropic · Claude apps
Smart reports (beta)
We’ve launched smart reports, which analyze how a team uses Claude and report on the work getting done, what it costs, where sessions run into friction, and which repeated patterns are worth packaging as shared skills. Smart reports are available in beta on Claude Enterprise plans. Learn more in Get started with smart reports.
OpenAI · Codex CLI
Codex CLI Cygwin
OpenAI · Codex CLI
Codex CLI Python
DeepSeek · New model
DeepSeek-V4.1-Flash Release
Today, we officially release the DeepSeek-V4.1-Flash model. It is the smallest model in our new architecture family, with native multimodal visual understanding. The new architecture is designed for a higher capability ceiling, faster inference, higher throughput, and scaling to larger models. * Tested only on the pure-text subset of the HLE benchmark set.
Google · Gemini app
Bring Google Gemini to your PC with the new Windows app
What: The Gemini app is now available on Windows. Press Alt + Space to summon Gemini instantly from anywhere on your PC whenever you need it. You can get quick answers, draft messages, and brainstorm ideas right alongside your active applications. Inside your dedicated desktop workspace, you have access to everything you love about Gemini. You can connect to Google apps like Gmail and Google Drive to summarize documents and find information quickly, generate custom images with Nano Banana, and dive into in-depth research on any topic. The app is available on Windows 10 or later. Download it today at gemini.google/desktop. Why: Your desktop is where you get work done. Bringing Gemini straight to your PC with instant shortcut access keeps you in your flow, making it simple to get answers and create visuals right alongside your everyday tools.
Google · Antigravity
Background service registration with remote-control, expanded half-page scrolling, and content safety…
Introduces remote-control service manager subcommands to run the CLI as a persistent background daemon, broadens half-page scrolling support and keybindings, and resolves issues with content filter handling, scratch directory watchers, plugin MCP initialization, schema validation, and conversation resumption.
Anthropic · Claude Code
Claude Code 2.1.267
3 new41 fixes9 otherAdded maxEffortLevel setting (top-level or per model under modelSettings): caps the effort level on every provider, including Bedrock, Vertex and Foundry; users can still pick a lower level. Added --system-prompt-snapshot off to render the system prompt fresh on every request instead of reusing the conversation's recorded prompt (for iterating on prompt text).
Anthropic · Claude API
For cache diagnostics, the API now stores a request’s fingerprint only when the request includes the…
A request that sends only the cache-diagnosis-2026-04-07 beta header is still accepted, but no fingerprint is stored. A later turn that points previous_message_id at it reports previous_message_not_found. Include diagnostics on every turn, with "previous_message_id": null on the first.
OpenAI · Codex CLI
Codex CLI 0.154.0
6 new6 fixes2 otherGPT-6-Astra is now available in the model picker and Amazon Bedrock catalogs. Experimental worktree support lets you create isolated checkouts for new or forked sessions using --worktree or /worktree, then browse and resume them.
Google · Antigravity
Model API retry resilience, headless print optimizations, and detailed tool approval prompts
Improves resilience to transient model API errors with extended exponential backoff, accelerates startup and sign-in by reading cached credentials, streamlines headless execution and error reporting, enhances tool approval prompts with contextual action descriptions and reasons, and fixes issues across plugin reinstalls, plugin MCP working directories, subagent lifecycle states, and command memory leaks.
Google · Antigravity
Introducing new Documents section and several UX enhancements for organization and control
External files such as Google Drive links, PDFs, and Office documents now have their own Documents section in the sidebar. This release also adds a virtualized viewer for large code and data artifacts, an option to hide whitespace-only changes in diffs, and clearer warnings when a configuration file fails to parse.
Anthropic · Claude Code
Claude Code 2.1.266
1 fixFixed a 2.1.265 regression affecting LLM-gateway and proxy setups: the undocumented CLAUDE_CODE_USE_GATEWAY environment variable, previously ignored unless ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN were both set, began forcing Cloud-gateway sign-in on its own in 2.1.265, so configurations that set it alongside an API key, apiKeyHelper, or custom auth headers failed every request with "Not signed in to the Cloud gateway". The variable on its own is ignored again; no configuration change is needed.
Google · Gemini CLI
Gemini CLI 0.59.0
2 fixes2 otherprevent SSRF in MCP OAuth metadata discovery and authentication. enforce fail-closed workspace trust and filter mcpServers in restricted mode.
Anthropic · Claude Code
Claude Code 2.1.265
4 new34 fixes12 otherAdded user.email and user.groups to the telemetry Claude Desktop and Cowork send through a Claude apps gateway, matching terminal sessions. Added support for pointing --plugin-dir at a folder of plugins: each child folder with a manifest loads, and children added or removed while running are picked up.
OpenAI · Codex app
Codex app
Added a configurable shortcut to open web links in your default browser. Added Open with > New tab to file diffs so you can open a source file at the selected line in the app. Performance improvements and bug fixes. Improved recovery and responsiveness when reopening chats or sending messages while a chat resumes. Fixed duplicate follow-up messages appearing in restored chat history.
OpenAI · Codex app
ChatGPT for iOS
Reference other tasks directly in the composer with @ mentions. Answer live questions while Codex continues working, without losing your draft. Explore files with Back navigation, recent files, and remembered reading positions. Choose a starting branch for new worktrees, or include your current local changes. Worktree setup can now continue in the background on iOS 26, with progress shown in a Live Activity. Voice now respects your selected model and reasoning effort. Restored missing delegated-task messages and fixed navigation between linked tasks. Code reviews now complete more reliably, without duplicate activity or stuck running states. Fixed crashes and stale queued messages when switching from Queue to Steer. Improved reconnect reliability and preserved device pairing during temporary service outages. Resolved duplicate plugin skills in composer suggestions.
Anthropic · Claude Code
Claude Code 2.1.263
1 otherBug fixes and reliability improvements.
OpenAI · Codex app
Codex MCP server removed
The codex mcp-server command and standalone codex-mcp-server binary have been removed after their deprecation on August 24, 2026. Update integrations that launch either command before upgrading Codex. Use the Codex app server for integrations. The app-server command is experimental and isn't supported for production workloads. Codex continues to support connecting to external MCP servers through codex mcp.
Google · Antigravity
Mid-conversation model execution, custom agent subagent dependencies, and MCP parameter validation fixes
Introduces /model for running one-off prompts with alternative models, adds conversation_title metadata to status line scripts, adds subagent dependency declarations in custom agent Markdown frontmatter, and delivers fixes for MCP argument schema validation, headless permission handling, Vim normal mode shortcuts, and cross-conversation artifact access prompts.