Gemini 3.1 Pro Preview carries forward the multimodal foundation of the Gemini 3 series as Google's frontier reasoning model, built to handle high-precision tasks across text, image, video, audio, and code within a one-million-token context window. The 3.1 update refines core intelligence with measurable gains on ARC-AGI-2, GPQA Diamond, and SWE-Bench Verified benchmarks, validating stronger performance in reasoning, scientific evaluation, and software engineering. It introduces a medium thinking level option that lets developers balance cost, speed, and depth more flexibly, and improves long-horizon stability and tool orchestration for complex agentic workflows. The model is designed to excel in autonomous coding, structured planning, and enterprise-grade workflow automation where consistent multi-step reasoning matters.

The model builds on Gemini 3 Pro with targeted improvements for production domains—particularly software engineering, financial modeling, and spreadsheet automation—where reliability and task continuity across extended contexts are essential. More efficient thinking reduces per-task token consumption compared to its predecessor without sacrificing output quality, which translates to lower operating costs for teams running long agentic sessions. According to the Artificial Analysis Intelligence Index, it leads that benchmark while commanding less than half the price of comparable frontier models, making it a practical choice for developers building code automation, financial analysis pipelines, or autonomous agent systems. It's well-suited for high-context enterprise tasks where multimodal input, structured output, and sustained reasoning across dozens of turns create real value.