GitHub made Gemini 4 Carbon, Google’s latest model, generally available in Copilot on November 10, 2026. The company positions it for complex coding, agentic work and reasoning, with a particular emphasis on secure code generation. It joins an already crowded model picker, available on the Pro+, Max, Business and Enterprise plans.
A model positioned around code security
In its official post, GitHub presents Gemini 4 Carbon as especially suited to security-related tasks: vulnerability detection, review of sensitive code, patch generation. No benchmark comparison is published against Gemini 4 Argon, Google’s previous model added to Copilot a month earlier, nor against the rest of the picker’s lineup. Billing follows the provider’s usage-based list price.
Plans and surfaces covered
The model is available on Copilot Pro+, Max, Business and Enterprise; Copilot Pro is excluded. It is rolling out gradually in the model picker across Visual Studio Code, Visual Studio, Copilot CLI, the coding agent, the Copilot app, github.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse. Because the rollout is staggered, some eligible accounts will not see it right away.
| Date | Model added to Copilot | Provider |
|---|---|---|
| September 4, 2026 | Claude Fable 5.1 | Anthropic |
| September 22, 2026 | Grok 4.7 | xAI |
| October 9, 2026 | Claude Haiku 5.5 | Anthropic |
| October 15, 2026 | Gemini 4 Argon | |
| November 10, 2026 | Gemini 4 Carbon |
Governance: what admins actually control
For Business and Enterprise accounts, access is governed through Copilot’s model policy. GitHub generalized the corresponding API in late October, which makes it possible to automate enabling or disabling a model per organization instead of relying on the admin UI for every release. By default, a new model is enabled automatically unless an admin has turned off global default enablement or blocked this specific model.
# Check an organization's Copilot seat and billing allocation
curl -H "Authorization: Bearer $GH_TOKEN" \
-H "Accept: application/vnd.github+json" \
https://api.github.com/orgs/ORG/copilot/billing
Gemini 4 Carbon joins a picker that has gained several new entries within a matter of weeks, Grok, GPT, Claude and Gemini alike, with none of them settling in as a lasting default.
A release cadence that is not slowing down
Gemini 4 Carbon arrives three days after GitHub gave Copilot agent-assisted pull request merging, and in the same week Anthropic unveiled Claude Fable 5.2 and Mythos 5.2, also built around agentic coding. For engineering teams, this sustained pace makes choosing between models harder: every release promises gains on reasoning or security, rarely backed by comparable public benchmarks.
Key takeaways
Gemini 4 Carbon has been generally available in GitHub Copilot since November 10, 2026, for the Pro+, Max, Business and Enterprise plans, with marketing centered on code security. Organization admins keep control through the model policy, which can be automated via API, and the rollout remains gradual across accounts.
The “built for security” pitch comes back with every new code model released in the past year, without any vendor publishing a verifiable testing methodology to back it up. Before moving an entire team onto Gemini 4 Carbon for security reviews, I recommend testing it against a sample of the project’s own known vulnerabilities, not against the press release’s promise. — Simon Janvier
To go further: the full announcement is available on GitHub’s official changelog.
