GPT-6 Astra lands in GitHub Copilot across VS Code, JetBrains, Xcode
On by default for Pro+ and up at provider list pricing — plus Copilot CLI model fallback, npm multi-config publishing, and Copilot PR approvals.
Copy markdown
Astra, everywhere you already code
GPT-6 Astra is GA across 10 Copilot surfaces — VS Code, Visual Studio, Copilot CLI, the coding agent, github.com, GitHub Mobile, JetBrains, Xcode, and Eclipse. It's on by default for Pro+, Max, Business, and Enterprise and billed at provider list pricing under usage-based billing, no premium multiplier. Rollout is gradual; Business/Enterprise admins can gate it via model policy.
Copilot CLI agents get model fallback
v1.0.83 lets a custom agent list several models in its `model` field and try each until one is available, with `model-policy: required` locking users to that list. It also adds Claude Fable 5.1 and, on Linux, restricts sandbox network egress to configured proxies.
npm: many trusted-publishing configs per package
You can now attach multiple OIDC configurations to a single package — separate paths for stable, prerelease, and staging — instead of the old one-config limit. A publish is authorized if the token matches any config, all support staging by default, and the versions tab shows approve/reject history. One less reason to keep long-lived npm tokens in CI.
Copilot can now approve pull requests
Copilot code review can submit formal PR approvals — off by default and configurable at the enterprise, org, and repo level, and scopable to specific file paths. Like a human reviewer's, its approval is dismissed the moment new commits land, so it never rubber-stamps stale code.
Elsewhere: Gemini gains agentic video understanding
Gemini 3.7/3.6 Flash and 3.5 Flash-Lite can now navigate a video's timeline on demand — pulling transcripts, frames, or audio only as needed — using up to 88% fewer tokens on long clips than static processing. Cheaper long-video pipelines, no extra plumbing.