Gemini 3.1 Pro: benchmarks, price and rank

Its headline results

The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.

• APEX-Agents35.1%

  Long tasks written by investment banking analysts, management consultants and corporate lawyers from their own work, done across files and apps in a realistic work setup.

  48th of 56. Independent result from Epoch AI.

• BrowseComp85.9%

  Questions whose answers are hidden deep on the web and take many searches and pages to track down.

  10th of 56. The maker's own number, from Google, via LLM Stats.

• MCP Atlas69.2%

  Multi-step tasks where the model has to pick and combine many real tools, connected through MCP, the standard way agents plug into apps and data.

  19th of 33. The maker's own number, from Google, via LLM Stats.

• SWE-Bench Pro54.2%

  Long changes to real codebases in several languages, often across many files, including code from private and copyleft projects that models are unlikely to have seen.

  37th of 50. The maker's own number, from Google, via LLM Stats.

• Humanity's Last Exam46.4%

  About 2,500 questions written by experts in more than a hundred fields to be as hard as they could make them, many needing a diagram or image.

  Independent result from Epoch AI.

• Vending-Bench 2$3,774

  The model runs a simulated vending machine business for a year on its own: finding suppliers, ordering stock, setting prices and paying the daily fees.

  43rd of 60. Independent result from Epoch AI.

• DeepSWE 1.112%

  Software engineering tasks run in the same simple agent setup for every model.

  Independent result from LLM Stats.

• SimpleBench79.6%

  Multiple-choice puzzles about everyday situations, space, time and people, written to trip up models that pattern-match instead of think. Ordinary adults do well on them.

  12th of 68. Independent result from Epoch AI.

• SimpleQA Verified73.5%

  Short factual questions with one checked answer, asked without web search.

  6th of 78. Independent result from Epoch AI.

• FrontierMath59.6%

  Hundreds of new, unpublished math problems written by mathematicians, from hard undergraduate level up to research, with answers a computer can check.

  42nd of 76. Independent result from Epoch AI.