Image: GitHub / github.blog GitHub Copilot Gets GPT-6 Astra and Gemini 3.8 Flash This Week
Two major model additions land in Copilot within 24 hours: GPT-6 Astra is generally available across all plans, and Google's new Gemini 3.8 Flash joins with discounted introductory rates through December.
GitHub Copilot had a busy week. Between September 3 and 4, it added two major models: Google’s Gemini 3.8 Flash and OpenAI’s GPT-6 Astra. Both are available now across all paid Copilot plans.
GPT-6 Astra in Copilot
GPT-6 Astra, which OpenAI released on September 3 with claims of representing a generational leap in AI capability, is generally available in GitHub Copilot as of September 4. That’s a fast turnaround, and GitHub is positioning it specifically for autonomous coding tasks.
This lands on the same day OpenAI is rolling out broader API access for Astra. Copilot users on Pro, Pro+, Max, Business, and Enterprise all have access through the model picker.
Gemini 3.8 Flash
Google’s new Flash model showed up in Copilot on September 3. GitHub’s characterization in the changelog: Gemini 3.8 Flash “performed strongly on complex terminal-based coding tasks and demonstrated rigorous validation and persistent recovery from actionable failures.”
That’s an interesting way to describe it. Recovery from failures in terminal tasks is a meaningful signal for developers who run Copilot against long build chains or CI scripts. The model is available in VS Code, Visual Studio, Copilot CLI, the GitHub Copilot cloud agent, the Copilot app, JetBrains IDEs, Xcode, and Eclipse.
Pricing uses introductory provider rates through December 31, 2026, under usage-based billing.
Code review can now approve PRs
A meaningful feature landed September 1: Copilot code review can now submit approvals on pull requests. Repository administrators have to enable this, and when they do, Copilot’s approval counts toward required-approval rules the same way a human reviewer’s would.
If new commits push after Copilot approves, the approval gets dismissed automatically, matching standard reviewer behavior.
This has been a long-requested feature. The catch is the admin gate, which means individual developers can’t enable it themselves. But for teams that can configure it, it removes a common bottleneck in smaller codebases where getting a second human approval takes time.
Model deprecations
Several models are being deprecated in Copilot on October 2, 2026. GitHub hasn’t published the complete list in the changelog summary, but the notice covers Copilot Chat, inline edits, ask and agent modes, and code completions.
Enterprise administrators can also now set a preferred model as the default for new conversations through managed settings, and content exclusion policies are now generally available in the Copilot app and CLI.
Sources: GitHub Changelog, Gemini 3.8 Flash in Copilot