Gemini 3.1 Pro reaches developers in preview

Google introduced Gemini 3.1 Pro on 19 February, bringing its revised reasoning model to consumer and developer products. Developers received preview access through the Gemini API in AI Studio, Gemini CLI, Antigravity and Android Studio. Enterprise routes included Vertex AI and Gemini Enterprise; consumer rollout included Gemini and NotebookLM.

The announcement reported a 77.1% result on ARC-AGI-2, more than double the company’s stated Gemini 3 Pro score. That measures one benchmark’s logic-pattern tasks, rather than proving the same improvement in every workplace task. The preview designation also matters for developers choosing production dependencies: access to a stronger model is distinct from a settled interface, pricing assumption or reliability record.