Google on February 19, 2026, released Gemini 3.1 Pro, an updated AI model the company says delivers a significant leap in complex reasoning, and began rolling it out across consumer, developer, and enterprise platforms the same day.
On the ARC-AGI-2 benchmark for abstract logic tasks, Gemini 3.1 Pro scores 77.1 percent, more than doubling its predecessor Gemini 3 Pro, which scored 31.1 percent. The Decoder Google says the model also outperforms competing offerings from Anthropic and OpenAI on that test, though benchmark comparisons in AI carry inherent limitations and should not be taken as a complete measure of real-world performance.
This is the first time Google has incremented a Gemini model by a 0.1 step; previous mid-cycle upgrades were 0.5 releases. Thurrott
The model is designed for tasks requiring advanced reasoning, such as synthesizing data, explaining complex topics, or supporting agentic workflows. Google Google describes it as the foundational intelligence behind Gemini 3 Deep Think, a more specialized model updated the previous week for science, research, and engineering use cases. With 3.1 Pro, Google says those capabilities are now available to a broader audience for everyday applications. The Decoder
Improvements in this release include better software engineering behavior, enhanced agentic capabilities in domains like finance and spreadsheet applications, more efficient token usage, and the addition of a MEDIUM thinking level parameter that gives developers more flexibility to balance cost, performance, and speed. Google
Developers can access the model in preview through the Gemini API, Google AI Studio, the Gemini CLI, the agent-based development platform Google Antigravity, and Android Studio. Enterprise access is available through Vertex AI and Gemini Enterprise. Consumers can use it in the Gemini app and in NotebookLM, with the latter restricted to Pro and Ultra subscribers. The model is also accessible via Microsoft services including GitHub Copilot, Visual Studio, and Visual Studio Code. The Decoder
API pricing scales by prompt length and matches Gemini 3 Pro’s existing rates. The Decoder
The release remains in preview. Google says it plans to continue refining the model based on user feedback, particularly around agentic workflows, before moving to general availability. The Decoder
The announcement comes amid an accelerating pace of AI model releases from major technology companies. Google CEO Sundar Pichai noted in the company’s Q4 2025 earnings release that its first-party models now process over 10 billion tokens per minute via API, and that the Gemini app has grown to over 750 million monthly active users.