GPT-6.2 Astra and GPT-6.2 Sol arrive in GitHub Copilot
GitHub has made OpenAI's two new agentic coding models generally available in Copilot, with usage-based billing at provider list prices.
On November 17, 2026, GitHub said OpenAI's GPT-6.2 Astra and GPT-6.2 Sol are now generally available in GitHub Copilot. Both models, referred to together as GPT-6.2 Astra and GPT-6.2 Sol, are aimed at agentic coding and code review, and GitHub says both show stronger coding judgment than earlier GPT-6 models. The GitHub changelog entry covers availability, billing and admin controls, and describes the improvements in qualitative terms rather than publishing benchmark tables.
What GitHub actually says
GitHub makes three concrete claims, all framed as early testing rather than published results:
- Both models caught significantly more real issues in code review than earlier GPT-6 models while using fewer tokens.
- GPT-6.2 Astra caught significantly more issues in code review than GPT-6 Astra and completed agentic coding tasks using noticeably fewer tokens.
- GPT-6.2 Sol caught nearly twice as many real issues in code review as GPT-6.1 Sol, while keeping its findings accurate.
Those claims matter because the two models are split by job. Astra is described as the model for agentic coding and terminal work. Sol is described as the model for complex agentic coding and in-depth code review. Neither is presented as a general chat upgrade; both are pitched at work where the model edits, runs and reviews code.
Who gets access, and where
Availability is limited to Copilot Pro+, Max, Business and Enterprise. The models appear in the model picker in Visual Studio Code, Visual Studio, Copilot CLI, the GitHub Copilot coding agent, the GitHub Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode and Eclipse.
Rollout is gradual, so not seeing the models yet does not mean you are ineligible. For Business and Enterprise, administrators manage access through the model policy in Copilot settings. Under default model enablement, new models switch on automatically unless an admin has turned off the global default or explicitly disabled these models. That means whether a given team can use them often depends on a setting someone else controls.
Billing is at provider list pricing under usage-based billing, with GitHub pointing to its models and pricing page for Copilot for the numbers.
What changes for a builder
Two things are worth watching, and they pull in different directions. The first is review quality. If Sol really catches close to twice as many real issues as GPT-6.1 Sol without losing accuracy, automated review becomes a stronger gate before a merge. The second is token efficiency. Agentic loops burn tokens on every step, so a model that finishes the same task with noticeably fewer tokens changes how much work an agent can complete inside a fixed budget.
Cost is the catch. Usage-based billing at list price means a stronger model can still cost more per call even when it uses fewer tokens per task. The changelog does not give per-token rates, so the net effect on a monthly bill cannot be worked out from this post alone.
Limits and open questions
There are no published benchmark scores, no context window figures and no per-token prices in the announcement. "Significantly more" and "nearly twice as many" are vendor characterisations without underlying data, and the testing conditions are not described. The post also does not say whether these models are reachable outside Copilot, for example through OpenAI's own API, or when the gradual rollout will finish. Independent evaluation of code review accuracy on real repositories is the missing piece.
What it means for building with AI
Model choice is now a routine part of the workflow rather than a one-time decision. A coding agent and a reviewer may want different models, and the same model can be cheap on one task and expensive on another. If you build apps by describing them and letting an AI write the code, the same logic applies at the tool level: Draftship's paid plans add a choice of AI model, and the docs on using the AI chat and choosing a model explain how that selection works. Draftship is not affiliated with GitHub or OpenAI, and nothing here suggests Draftship supports these specific models.
Primary sources
- OpenAI’s GPT-6.2 Astra and GPT-6.2 Sol now available · GitHub Changelog, Nov 17, 2026
This article was drafted with AI assistance from the primary sources listed below and checked automatically against them. An editor hasn't reviewed it yet.