GitHub announced on 17 November 2026 that OpenAI’s GPT-6.2 Astra and GPT-6.2 Sol are generally available in Copilot. Both are pitched at agentic coding (AI that plans and carries out multi-step tasks) and at code review.
What GitHub says
- Astra is aimed at agentic coding and terminal work. In early testing GitHub reports it caught significantly more code review issues than GPT-6 Astra and finished agentic tasks with noticeably fewer tokens.
- Sol is aimed at complex agentic coding and in-depth review. GitHub reports it caught nearly twice as many real issues as GPT-6.1 Sol while keeping findings accurate.
- Plans. Copilot Pro+, Max, Business and Enterprise.
- Where. VS Code, Visual Studio, Copilot CLI, the coding agent, the Copilot app, github.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse, with a gradual rollout.
- Billing. Usage-based, at provider list pricing.
- Admin control. Business and Enterprise admins manage access through the model policy. Under default model enablement, new models switch on automatically unless an admin has turned that off.
These are GitHub’s early-test figures, not independent benchmarks.
Why it matters
Notice where the marketing has moved. The pitch is no longer only “writes code faster”. It is “finds more real problems” and “uses fewer tokens”. Review is where AI-assisted projects either keep their savings or lose them, and tokens are now a line on the bill.
So what
If a supplier says they use AI to build your software, ask which models they use for building and which for review, who pays for usage, and how a human signs off. Our AI-assisted development work keeps those answers in the open. To scope a project, start a project.