Skip to content

Dataset · gaia-benchmark

GAIA Benchmark for Agentic LLMs

gaia-benchmark/GAIA

450+ non-trivial questions with single-answer targets for testing tool-using and search-equipped LLMs.

What's inside

GAIA is a benchmark for evaluating next-generation LLMs with access to tools, search, and autonomous actions. It features over 450 non-trivial questions with single-answer targets requiring varying levels of tool use and autonomy. Data includes audio, documents, images, and text in English; access is restricted for anti-scraping protection.

What you can do with it

  • Evaluating agentic LLMs with tools
  • Testing multimodal QA
  • Comparing models on the GAIA leaderboard
Languages
English
Size
up to 1K rows
Downloads this month
26.9K
Likes
999
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("gaia-benchmark/GAIA")