Gemini 3.1 Pro

Gemini 3.1 Pro, released by Google DeepMind in February 2026, represents a major leap forward in capabilities, topping 13 out of 16 major industry benchmarks. The model achieves 77.1% on ARC-AGI-2 (a benchmark measuring abstract reasoning and general intelligence), 94.3% on GPQA Diamond (graduate-level science questions), and a LiveCodeBench Pro Elo rating of 2887, demonstrating exceptional performance across reasoning, science, and coding domains.

On specialized benchmarks, Gemini 3.1 Pro scores 85.9% on BrowseComp (web browsing and information synthesis) and 69.2% on MCP Atlas (multi-modal context processing). These results highlight the model's strength not just in traditional NLP tasks but also in complex, real-world scenarios requiring tool use and multi-step reasoning.

A notable feature is the "medium" parameter that allows users to trade off between compute intensity and latency, providing flexibility for different application requirements. Gemini 3.1 Pro's broad benchmark dominance and configurable performance characteristics make it a strong contender in the most advanced tier of language models, competing directly with GPT-5.2, Claude Opus 4.6, and other frontier models.