Each week's verified changes, Act now first. Atom feed, one entry a week.
7 act now, 15 new capabilities
Since November 6, 2026, requests to gemini-embedding-2-preview route to gemini-embedding-2 and are billed at Gemini Embedding 2 rates. Update the model string; the API schema, pricing and embedding space are the same, so no other code changes or re-indexing are needed.
With telemetry or feature flags off, including behind LLM gateways, auto mode now asks about or denies calls the server can't check, with no fallback to the built-in classifier. Set CLAUDE_CODE_AUTO_MODE_SERVER=0 to keep using it. Bedrock, Vertex and Foundry are unchanged.
From December 3, 2026, API requests to gemini-omni-1.1-flash automatically route to gemini-omni-1.2-flash and are billed at Gemini Omni 1.2 Flash rates.
Claude Code no longer installs or updates the IDE extension when started in an IDE terminal; use /ide from an outside terminal or the editor's extensions view. The self-hosted runner orchestrator's log lines change: update scripts that match retryable=true: or retryable=false:.
The Install Extension and Run VS Code Command agent tools, the Create New Workspace tool and the /new slash command are no longer available in Local agent sessions. Remove #installExtension and #runCommand references from custom agents and prompt files.
From December 1, 2026, gemini-2.5-pro calls route to gemini-3.8-flash and are billed at its rates. Breaking changes include function calling, thought_signature validation and thinking_level replacing thinking_budget. Update the model string first.
From December 1, 2026, gemini-2.5-flash calls route to gemini-3.5-flash-lite and are billed at its rates, with the same breaking API changes as the 3.8 Flash migration. Update the model string first.
From a Slack or Teams conversation, @GitHub can investigate bugs, create issues, implement changes and open pull requests. Teams adds Canvas to review implementation plans, code changes and previews; issues and pull requests it creates now name the conversation they came from.
Admins can read and update Copilot code review settings through the APIs, from enterprise policy down to individual repositories. Enterprise and organization endpoints don't support GitHub Apps or fine-grained personal access tokens. Code review agent sessions now show the AI credits consumed.
Organizations can set an IP allowlist in the Claude Console, in beta. Requests to covered services from addresses not on the list are refused, which limits where a leaked API key works. It takes IPv4 and IPv6 addresses or CIDR ranges, up to 1,000 per organization.
gpt-6-cyber helps investigate vulnerabilities, reproduce them in an authorized test environment and verify fixes. Organizations need Daybreak Red access and approval; OpenAI's Daybreak setup guide covers what to do before the first request.
gemini-omni-1.2-flash generates video up to 15 seconds with conversational editing, improves character and subject consistency across up to 6 reference images and doubles paid-tier rate limits. Video output is priced at $12.00 per 1M tokens, 30% lower.
The API now accepts PPT, ODP and ODS input and adds page-image processing for DOC files. Function responses in the GenerateContent and Interactions APIs can return audio, video and Office or OpenDocument files, inline or by URI.
On paid plans with cloud agents enabled, an agent can scaffold a web app, spin up a Postgres database, put it behind Cursor SSO and host it on a domain you pick. Nothing goes live until security checks pass and you publish; team admins can approve or unpublish apps.
Copilot now works inside Xcode's own coding assistant, set up under Xcode > Settings > Intelligence. It starts with Xcode 27.3 beta, needs macOS Tahoe 26.6 or later and supports all Copilot plans. The separate Copilot for Xcode extension remains available.
gpt-image-2.6-sunburst and gpt-image-2.6-flare work in v1/images/generations and v1/images/edits, with adjustable complexity, SVG generation and animated GIF generation. Markup-guided editing is added for supported GPT Image 2.5 and 2.6 models.
Adds a cost report endpoint to the Claude apps gateway admin API, a warning when /schedule gives a cloud task connectors you did not request, a prompt before tools send your account email, and onFailure: block for Agent SDK hooks.
You can delegate coding tasks to Copilot, guide them and review the results inside JetBrains Air, through the Agent Client Protocol (ACP).
Model Vault now runs open models from DeepSeek, Qwen, Gemma, GPT-OSS, Mistral and others on dedicated infrastructure, with zero data retention by default and serving in the US, EU and Canada.
Text these models generate carries Anthropic's text watermark on the platforms Anthropic lists. On Claude Sonnet 4.6, text from a request that sets temperature, top_p or top_k to a non-default value might not carry it.
With the message-threads-2026-08-12 beta header the API keeps the conversation server-side, so later requests send only new messages. A thread expires 5 minutes after its last request.
Every organization can now connect GPT-Live (gpt-live-1, v1/live/sessions) to a phone number through its SIP provider for outbound calls.
5 act now, 2 pricing changes, 9 new capabilities
Fast mode requests on Opus 4.8 will error after November 30, 2026, with no fallback to standard Opus 4.8. The release also adds CLAUDE_CODE_SUBAGENT_EFFORT and pooled group budgets for the Claude apps gateway.
After November 30, 2026, requests to claude-opus-4-8 with speed: fast return an error. Move to Claude Opus 5.5 to keep fast mode.
In the Agents API public beta, context compaction now runs in the background by default and may increase token usage; set compaction.async to false for blocking compaction. Sessions can also update their tools: an update replaces the full tool list.
The Usage and Cost API and the usage and cost report endpoints of the Claude Enterprise Analytics API are now legacy but still supported. Their replacement, the Usage and Cost Analytics API (beta header analytics-usage-cost-2026-08-24), reports token usage and cost by model, workspace, API key or user; Anthropic publishes migration guides.
New paid-tier projects using generateContent, streamGenerateContent and batchGenerateContent store requests by default (store=true), like the Interactions API; existing projects keep their setting. Opt out with store=false per request or by turning off Generate Content API logging in the Logs and Datasets settings.
Through a GitHub or Microsoft account team, GitHub Enterprise Cloud can include Copilot for every licensed user, with AI usage metered from the first request against the enterprise budget and no monthly AI credits per seat. Users without an Enterprise license get Copilot Standalone at $10 per user per month; self-serve comes later.
Eligible ChatGPT Pro and Business subscribers can claim up to $500 per month in free OpenAI API credits as a subscription benefit, activated from ChatGPT or the API Platform by linking an API organization they own.
gpt-6-rosalind adds specialized biological reasoning with scientific tool use. Only organizations approved for Rosalind Research through the trusted-access program can use it.
The Computer Use tool is now GA in both the Interactions and GenerateContent APIs.
Copilot Fusion replaces picking a model: it routes and combines models per prompt at a chosen effort level. Each model it uses is billed at its standard rate, with a limited-time 30% discount on paid plans. Earlier auto, HydraFusion or HydraFusion Max selections move to the matching Fusion effort levels.
Copilot chat in VS Code, the Copilot app and Copilot CLI can now generate and edit images with GPT Image 2.5 Flare, saving them to the workspace. In public preview for Pro, Pro+, Business and Enterprise; billed separately from the chat model in image tokens that count toward the plan's AI credits. Admins can disable it via model policies.
The maximum request size for the Messages API and the Token Counting API rose from 32 MB to 128 MB on the first-party Claude API.
VS Code 1.144 adds an Agent Host that keeps agent sessions running locally or remotely, with continuity across VS Code, Copilot CLI, the Copilot app and github.com. Agents use the Copilot harness by default; the Claude Code or Codex harnesses can be used instead with a Copilot subscription.
Google released gemini-api, a stateless, zero-dependency command-line client for AI coding agents, scripts and CI pipelines. It works on the Interactions API to generate text and media, analyze files and run Gemini managed agents.
OpenAI's GPT-6.1 Terra, aimed at everyday agentic coding and terminal work at a lower cost than GPT-6.1 Sol, is GA for Copilot Pro, Pro+, Max, Business and Enterprise. It is billed at provider list pricing under usage-based billing; admins control it through the model policy, and new models are on by default.
gpt-6.1-terra is a fast, efficient model for visual reasoning, precise computer use and clear writing, available in v1/responses and v1/chat/completions; OpenAI's GPT-6 guide covers prompting and migration.
5 act now, 1 pricing changes, 6 new capabilities
The legacy GitHub app in Teams will be renamed GitHub (deprecated) and retired on November 16. Install the consolidated GitHub app, which adds Copilot, to keep using Copilot in Teams; it can turn an issue notification into a pull request via @github.
From October 22 to 26, 2026, GitHub rolled out enforcement of the global default policy on Copilot Business and Enterprise: eligible GA features and MCP-server settings left Unconfigured are set to the chosen default (Enabled, Disabled or Let organizations decide). Explicit choices are kept; review AI Controls → Copilot → Default policy for new features.
Self-hosted runner checkout hooks are now stopped after 30 minutes by default, failing the session's setup; raise --checkout-hook-timeout-min or set it to 0 if yours need longer. The release also adds /claude-api api-version to move code to Claude API version 2026-10-21.
Anthropic released API version 2026-10-21 and new SDK and ant CLI versions that send it by default (Python 2.0.0, TypeScript 1.0.0, C# 13.0.0, Go 2.0.0, Java 3.0.0, PHP 1.0.0, Ruby 2.0.0, CLI 2.0.0). The SDKs also carry breaking changes unrelated to the API version and move from MIT to Apache 2.0.
Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini and Grok 4.5 left most Copilot experiences on 2026-10-19. Suggested alternatives: Gemini 3.8 Flash, GPT-6.1 Sol, GPT-6 Luna and Grok 4.7; Enterprise admins may need to enable them in model policies.
In the Responses API you can choose 12-hour prompt caching, which keeps shared instructions, tools and context cached for hours so later requests get cached-input pricing; 12-hour cache writes cost 2.5x the input token price.
Claude Models on Google Cloud (Anthropic managed) is the Claude API run by Anthropic and reached through your Google Cloud project: Google Cloud IAM controls access and Google Cloud Marketplace handles billing. Fast mode for Claude Opus 5.5 is available there too.
The Decisions API, which turns text and images into typed answers, is GA with gpt-6-luna. It adds prompt-cache discounts for shared input, images in per-question instructions, structured input_json, and up to 400 questions per request (200 in the beta).
Two reranking models: Pro for reasoning-heavy and domain-specific queries, Fast for low latency. Adds Context Focus (returns the most relevant passages as scored spans in the same call), calibrated relevance scores, 100+ languages and a 32k-token context window.
If you run a platform or gateway, or resell API access, you can tell Anthropic which end customer each request is for, so a safety enforcement action can apply to that customer instead of your whole organization. Since 21 Oct, profiles also show enforcement_status and enforcement_actions.
Cursor's Agent SDK defines an agent as a folder of code (instructions, typed tools, skills, MCP connections, channels), tests it with evals and a local playground, and deploys it to Cursor with agent-sdk deploy or to your own infrastructure. Agents wake on GitHub, GitLab, Bitbucket and Slack events.
GET and PATCH endpoints now read and update Copilot model policies for enterprises, enterprise teams, agent apps and organizations, so model availability can be managed as code. Enterprise policies can delegate model control to organizations or to teams and apps.
2 act now, 11 new capabilities
The release adds a notice to /design-sync and /design-login that standalone Claude Design (claude.ai/design) closes by 2026-12-14.
GitHub will deprecate Claude Opus 4.8*, Claude Opus 5 and GPT-5.6 Terra across Copilot experiences on 2026-11-16; suggested alternatives are Claude Opus 5.5, or GPT-6.1 Sol / GPT-6 Luna. *Opus 4.8 exception for individual annual-plan subscribers — see the official post.
New env var chooses the model the client-side auto-mode classifier uses, and /config gains an 'Idle compaction' row to turn idle compaction off or back on.
Create a custom voice from a text description via the API or voice Playground and reuse it in GPT-Live sessions; creating a prompt-generated voice is free, speech generation uses regular GPT-Live session billing. Nine new GPT-Live voices were also added.
Cloud agents can build interactive canvases you can share by link. The Environments API (beta) lets you programmatically create/manage cloud environments, which now support up to 300 repositories and load large environments 5x faster.
Requests to Claude Opus 5.5 no longer require access approval or the fast-mode-2026-02-01 beta header; a default fast-mode rate limit applies in Start, Build and Scale tiers. Fast mode for Opus 5 / 4.8 remains in research preview.
Agentic autofix remediates code-scanning alerts by exploring relevant files and proposing a fix, now generally available (per the GitHub Changelog Copilot feed).
You can trace Messages API requests and Claude Managed Agents sessions in your observability backend; the Claude SDK emits spans in Python, TypeScript, C#, Go, Java, PHP and Ruby after adding an OpenTelemetry destination in the Console.
Stable 1.142 (released Oct 14, 2026) adds preview shared canvases beside chat, lets you move a chat to another folder or Git worktree without losing history, expands the Markdown editor, and progressively makes the Copilot harness the default for new sessions.
Origin integrates with 20+ apps across deployments, CI, code review, security, monitoring and project management (including Cloudflare, Linear, Sentry and Socket Security). Any team can now submit an app for review.
Google's Gemini 4 Argon, designed for complex coding, agentic and reasoning tasks, is now generally available in GitHub Copilot.
Google released gemini-4-argon, described as its most capable model for complex coding, agentic, math and science tasks, with API changes, pricing and migration guidance in the linked 'What's new' page.
Added support for connecting to MCP servers on protocol revision 2026-07-28 (server/discover, no initialize). Enterprise admins can also set an Extension marketplace URL that becomes the only extension source.
6 act now, 4 pricing changes, 5 new capabilities
GitHub Copilot has ended support for JetBrains IDE 2025.1; the plugin now needs JetBrains IDE 2025.2 or later. The same update lets enterprise admins set the default agent model for new conversations through managed settings.
gemini-3.7-flash is deprecated, replaced by gemini-3.8-flash, and API requests are automatically routed to it. gemini-3.5-flash is likewise deprecated and auto-routed to gemini-3.6-flash. Google advises updating model strings explicitly.
That Deep Research agent version is deprecated and will be shut down on 2026-10-23. Migrate the agent parameter to deep-research-preview-04-2026 or deep-research-max-preview-04-2026.
claude-haiku-5-5 has a 1M-token context, 128k max output and adaptive thinking on by default. On it, budget_tokens thinking, temperature/top_p/top_k, assistant prefill and computer_20250124 can return 400 errors, as can resending thinking blocks after changing system, tools or earlier turns. Check Anthropic's migration guide.
To reduce data-exfiltration risk, web_fetch in Claude Managed Agents only fetches URLs that appeared earlier in the session (a user message, a web_search result or a fetched page). Others, such as URLs in Claude's output, the system prompt or tool output, return url_not_in_prior_context; send them in a user.message instead.
With limited networking, web_search and web_fetch now follow allowed_hosts (fetches elsewhere return url_not_allowed), and creating or updating a session fails with a 400 error when a web tool's allowed_domains has an entry outside allowed_hosts. Add the host to allowed_hosts or drop the entry.
Organization owners can bill Copilot code reviews from licensed members to the organization that owns the repository (needs AI Credits paid usage enabled, optional budget) instead of the member's own entitlement, where an exhausted quota makes the review fail. Owners and repo admins can also stop externally licensed users from requesting reviews.
Anthropic lowered the prompt-cache read price for Claude Sonnet 5.5 to 0.05x the base input price (from 0.1x). Cache writes and all other prices are unchanged.
Anthropic's Claude Max and Team subscription plans now include monthly Claude API credits; the documentation page on API credits for Max and Team plans explains how to claim them.
Usage tiers are now Build, Launch and Grow. Organisations automatically upgrade as total credit purchases reach tier minimums; monthly usage limits and per-model rate limits are documented on the usage-tiers page.
Use gpt-6.1-sol with service_tier: 'ultrafast' to reduce time between output tokens. Available to all API users subject to rate limits, with global processing and US/EU data residency; OpenAI points to its Ultrafast pricing page for cost.
Claude Haiku 5.5 (claude-haiku-5-5) is added and becomes the default Haiku model on the Anthropic API, with a 1M context at $0.10/$0.50 per MTok ($0.50/$2.50 for prompts over 100K).
See and reply to local agents running on your computer from the Cursor iOS app. On by default except for Enterprise organisations, whose admins enable it in Org settings > Security and identity.
gemini-nano-banana-2.1 is GA for image generation and conversational editing at Flash-level speed/cost, with 1:4 to 8:1 aspect ratios at 1K/2K/4K. The older gemini-3.1-flash-image is deprecated (no shutdown date announced).
mistral-large-4 is an open-weight (weights coming soon) multimodal MoE model with a 1M context window, now in Public Preview. Mistral lists launch pricing of 50% off for 2 weeks.
3 act now, 1 pricing changes, 4 new capabilities
Anthropic announced the deprecation of claude-sonnet-4-5-20250929, with retirement on the Claude API scheduled for 2026-11-30, and recommends migrating to Claude Sonnet 5.5.
mistral-ocr-4-0 retires on 2026-09-30; use mistral-ocr-4-1 or mistral-ocr-latest instead, at the same price. (The same entry also deprecates Leanstral 1.5 and Z.ai GLM 5.2 with later retirement dates.)
Launch of claude-sonnet-5-5. Breaking notes include: forced tool use (tool_choice any/tool) returns 400, and to turn off up-front thinking you send thinking {type: 'between_tools'} instead of 'disabled' at high effort or below. See Anthropic's migration guide before swapping model strings.
You can sign in to Devin with ChatGPT and use an eligible ChatGPT Plus or Pro plan for GPT model usage in Devin Desktop; those requests are billed to your ChatGPT plan.
Embed 5 supports text, image and mixed inputs in one vector, 100+ languages, 128k context, and Matryoshka dimensions from 256 to 2048. Pro and Fast share an embedding space, so a corpus indexed with one can be queried with the other.
gpt-6.1-sol for complex coding and professional work at a lower cost than GPT-6 Astra. Standard pricing quoted for prompts up to 272K input tokens: $2 input, $0.10 cached input, $2.50 cache write, $10 output per 1M tokens.
Agents in the Agents API can complete tasks in an OpenAI-hosted browser, with website access approvals and sign-in handled by your application.
Mistral lists Z.ai GLM 5.3 (zai-glm-5-3) as Generally Available, positioned as the replacement for the deprecated GLM 5.2 at the same price.
2 act now, 1 pricing changes, 1 new capabilities
The filename and title fields on file, project document and artifact activities are now always empty or omitted, including for past activities. To look up a name or title, use a Compliance Access Key with the read:compliance_user_data scope.
claude-opus-5-5: 1M-token context by default, 128k max output, always-on adaptive thinking, priced $4 input / $20 output per MTok (Opus 5 is $5/$25). thinking {type:'disabled'} or {'enabled'} returns a 400 error; omit thinking and control depth with effort.
Refusals before any output are billed again when stop_details.category is bio, frontier_llm or reasoning_extraction, at the rates of the model that ran. Mid-stream refusals were already billed; early refusals in other categories stay free, and fallback credit is unchanged. Applies on all platforms.
New reasoning models accepting text and image inputs. Standard pricing quoted for prompts up to 272K input tokens: Sol $2 / $0.20 cached / $10 output, Luna $0.10 / $0.01 cached / $0.50 output per 1M tokens.
2 act now
Google now limits access to the Gemini 2.5 models to users who have actively used them before. They are not deprecated and are still served through the API, but new projects should use 3.5 Flash-Lite or 3.8 Flash.
antigravity-preview-09-2026 replaces the deprecated 05-2026 agent. Remote-sandbox users who read only output_text or model_output just change the agent string; local tool runners and function_call parsers must adapt, because built-in tools were renamed, take PascalCase parameters and edit files by line range.