Copilot Business admins have to flip a policy before Gemini 3.1 Pro shows up in the picker. Google's 19 February 2026 blog puts the model in preview across the Gemini app, Vertex, and developer tools. Preview is not general availability.
The news
The Google blog "Gemini 3.1 Pro: A smarter model for your most complex tasks," dated 19 February 2026, says 3.1 Pro is rolling out that day: for developers in preview via the Gemini API in Google AI Studio, Gemini CLI, Google Antigravity, and Android Studio; for enterprises in Vertex AI and Gemini Enterprise; for consumers via the Gemini app and NotebookLM.
The same post states that on ARC-AGI-2, "a benchmark that evaluates a model's ability to solve entirely new logic patterns, 3.1 Pro achieved a verified score of 77.1%. This is more than double the reasoning performance of 3 Pro." The Frontier did not retrieve an independent ARC Prize leaderboard screenshot for that score. Treat 77.1 percent as Google's reported figure.
The post says 3.1 Pro in the Gemini app is rolling out with higher limits for Google AI Pro and Ultra plans, and that 3.1 Pro is available on NotebookLM exclusively for Pro and Ultra users. It says Google is releasing 3.1 Pro in preview "to validate these updates and continue to make further advancements in areas such as ambitious agentic workflows before we make it generally available soon." The GA date is UNKNOWN.
A Google Cloud blog the same day, by Michael Gerstenhaber, VP Product Management, Gemini Enterprise, repeats preview availability in Vertex AI, Gemini Enterprise, AI Studio, Android Studio, Antigravity, and Gemini CLI. It quotes JetBrains' Vladislav Tankov claiming "up to 15% improvement over the best Gemini 3 Pro Preview runs" in JetBrains evaluations. That is a named-customer quote on Google's page, not a public eval harness.
GitHub's changelog dated 19 February 2026 says Gemini 3.1 Pro is rolling out in GitHub Copilot public preview for Copilot Pro, Pro+, Business, and Enterprise, in VS Code (chat, ask, edit, agent), Visual Studio (agent, ask), github.com, and GitHub Mobile. Copilot Enterprise and Business administrators "must enable the Gemini 3.1 Pro policy in Copilot settings." Rollout is gradual.
Who is bound
Google shipped the model and the consumer/developer posts. GitHub shipped the Copilot picker listing. Copilot Business and Enterprise admins are bound by the policy toggle before staff see the model. Vertex and Gemini Enterprise customers consume preview terms, not GA SLAs, unless their contract says otherwise. The contract text is UNKNOWN.
What's new
Google frames 3.1 Pro as a 0.1 increment on Gemini 3 with Deep Think-class reasoning in the default product path, and as preview rather than a silent GA. The operator change that is not a benchmark is Copilot: a named model in the picker behind an admin policy. Vertex preview is the enterprise path for teams that already standardized on Google Cloud.
What it does not settle
Date: 19 February 2026. Status: preview; GA UNKNOWN. Price: Google's 19 February blog does not state API rates. Region: Gemini app and NotebookLM Pro/Ultra limits as stated; Vertex and Copilot availability follow those products' regions. Full region matrix UNKNOWN from these posts. Hardware: none. Data: ARC-AGI-2 77.1 percent is Google-reported. License: proprietary Gemini terms, not an open-weight drop.
What to do now
Copilot Business and Enterprise admins should decide the Gemini 3.1 Pro policy before developers assume the picker is on. Eval teams should rerun their own agent and coding suites. Do not file 77.1 percent ARC-AGI-2 as an internal score.
When preview becomes GA is UNKNOWN.