Gemini 3.1 Pro

Google DeepMind🇺🇸 United States
active

Google's most capable Gemini model. Next-generation reasoning with thinking mode and 1M context.

Context window1000K tokens
Input / 1M tokens$2
Output / 1M tokens$12

Version History

3.1minor

Gemini 3.1 Pro enters public preview in GitHub Copilot with focus on efficient edit-then-test loops and agentic coding capabilities.

gemini-3.1-pro-preview-0214major

Gemini 3.1 Pro released as upgrade to Gemini 3 Pro. Enhanced reasoning for complex multi-step problems. Preview access via AI Studio.

Benchmark Scores

Full leaderboard →
90.4%
AIME 2024
93.0%
AIME 2025
96.8%
DocVQA
94.3%
GPQA
95.8%
HumanEval
78.4%
LiveCodeBench
97.1%
MATH
92.6%
MMLU
91.4%
MMLU-Pro
85.1%
MMMU
80.6%
SWE-bench Verified

Coverage