GAIA: A Benchmark for General AI Assistants
GAIA is a benchmark of 450+ real-world questions requiring multi-step reasoning, web browsing, multi-modality handling, and tool use, designed to be easy for humans (~92%) but hard for AI assistants, scored across three difficulty levels.
What this benchmark measures
GAIA is a benchmark of 450+ real-world questions requiring multi-step reasoning, web browsing, multi-modality handling, and tool use, designed to be easy for humans (~92%) but hard for AI assistants, scored across three difficulty levels.
Rows on this page are sourced from public benchmark artifacts, leaderboard exports, or source-linked model reports. Each row keeps benchmark version, source model name, and available run details attached to the score.
The metric shown here is accuracy. It should be interpreted within GAIA: A Benchmark for General AI Assistants, not compared as part of a site-wide ranking.
Frequently asked
What is GAIA: A Benchmark for General AI Assistants?
GAIA is a benchmark of 450+ real-world questions requiring multi-step reasoning, web browsing, multi-modality handling, and tool use, designed to be easy for humans (~92%) but hard for AI assistants, scored across three difficulty levels. It is a agents benchmark measured by accuracy.
What does accuracy mean on GAIA: A Benchmark for General AI Assistants?
GAIA: A Benchmark for General AI Assistants reports accuracy (%); higher is better. Scores are shown only within GAIA: A Benchmark for General AI Assistants and are never averaged with other benchmarks.
What is the top reported GAIA: A Benchmark for General AI Assistants score?
Claude Sonnet 4.5 has the top reported score on GAIA: A Benchmark for General AI Assistants: 74.55% (accuracy).
Why do GAIA: A Benchmark for General AI Assistants scores differ across runs?
Harness, scaffold, reasoning effort, and prompt setup change results, so two runs of the same model can differ. evals.report keeps each score with its run context so the differences stay visible.
Does evals.report rank models across benchmarks?
No. GAIA: A Benchmark for General AI Assistants scores are shown within their own metric; evals.report never combines benchmarks into a composite ranking or a single "best model".