Gemini 3.6 Flash lands in GitHub Copilot across all editors
Google's cheapest agentic model enters the Copilot picker with top computer-use scores; GitHub Code Quality goes paid and Claude retires prompt APIs.
Copy markdown
Gemini 3.6 Flash, now in the Copilot model picker
Google's newest Flash model is now selectable in GitHub Copilot across VS Code, Visual Studio, JetBrains, Xcode, Eclipse, the Copilot CLI, and cloud agents — on Pro, Pro+, Max, Business, and Enterprise. In GitHub's testing it lifts DeepSWE from 3.5 Flash's 37% to 49% and tops OSWorld-Verified computer use at 83.0%, ahead of GPT-5.6 Luna and Grok 4.5, all at Google list pricing under usage-based billing. Business and Enterprise admins have to switch on the Gemini 3.6 Flash Preview policy before the model appears.
GitHub Code Quality is GA — and now metered
GitHub Code Quality left preview on July 20 and started billing the same day: $10 per active committer per month plus metered charges for its AI-assisted detection and Copilot Autofix, on top of CodeQL compute on Actions. It pairs deterministic CodeQL with AI to flag maintainability and reliability issues in pull requests — built for reviewing AI-written code — and is live on Enterprise Cloud and Team, but not Enterprise Server yet. Audit which repos have it switched on before the first invoice lands.
Claude's experimental prompt APIs retire August 17
Anthropic is sunsetting the legacy Workbench and its experimental prompt endpoints — /v1/experimental/generate_prompt, improve_prompt, and templatize_prompt — on August 17. If any of your tooling calls those to auto-generate or rewrite prompts, migrate now; the newer prompt tools in the Claude Console stay.