Frontier AI
Copilot Chat, agent mode and code completions drop GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, Gemini 3.7 Flash and Grok 4.5 on October 19.
Why it matters. Enterprise admins may need to enable replacements such as GPT-6.1 Sol and GPT-6 Luna in model policies before users see them.
Source: GitHub changelog
Claude models on Amazon Bedrock return JSON that matches your schema on bedrock-runtime and bedrock-mantle, through output_config.format or strict tool use.
Why it matters. Bedrock extraction and agent pipelines can drop custom parsing and retries, though endpoint and Region support varies by model.
Source: Amazon Bedrock User Guide
Generally available GET and PATCH endpoints read and set Copilot model policies for enterprises, enterprise teams, enterprise apps and organizations.
Why it matters. Admins can manage model availability in code, and enterprise policies can delegate that control to organizations, teams or apps.
Source: GitHub changelog
The new /auto-mode-setup command reads your project and recent sessions to configure auto mode, and works on Bedrock, Vertex, Foundry and gateway sessions.
Why it matters. Teams that run Claude Code through a cloud provider can now teach auto mode about their environment with a guided flow.
Source: Claude Code releases
Foundry agents get a managed browser on Playwright Workspaces to navigate sites, inspect content, fill forms and run multistep workflows from natural-language instructions.
Why it matters. Agents can now reach portals that lack APIs, with Live View, human takeover and approvals before sensitive actions.
Source: Microsoft Foundry blog
Bedrock now serves SpaceXAI Grok 4.7, built for coding, agentic tasks and knowledge work, in AWS GovCloud (US) through the console and APIs.
Why it matters. Government workloads that must stay in GovCloud can now use Grok 4.7 for repo-scale coding and browser agents.
Source: AWS What's New
A retrieval instruction sets search guidance, persona, answer format and answer policy for AgenticRetrieveStream requests, after Amazon Bedrock reviews it.
Why it matters. Knowledge base teams can pin an immutable version, test changes on DRAFT, and roll back by switching version numbers.
Source: Amazon Bedrock User Guide
Clairic, at @cf/cloudflare/clairic, extracts, computes, estimates and forecasts values from unstructured input, and bills input tokens only.
Why it matters. Cloudflare claims lower cost and higher accuracy than LLMs for extraction, and says open weights will follow.
Source: Cloudflare blog