Skip to content

Benchmark

GAIA

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models measured
71
Scores recorded
291
Sources
3
IQ weight
1.01

Score distribution

37%
Mean
32.2%
Median
28.6%
Std dev
2.3%
10th percentile
80.6%
90th percentile
0%
Min
0%94.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.01 for chat models, measured over 71 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
201GLM 5.3 FlashLevel 3· GAIA Leaderboard data16.3%
202qwen-coderLevel 2· GAIA Leaderboard data16.3%
203Claude Sonnet 4.5· GAIA Leaderboard data15.9%
204Meta Llama 3 70B Instruct TurboLevel 2· GAIA Leaderboard data15.7%
205GPT-4 TurboLevel 3· GAIA Leaderboard data15.4%
206GPT 4Level 1· GAIA Leaderboard data15.1%
207GPT 5.1Level 2· GAIA Leaderboard data15.1%
208Meta Llama 3 70B Instruct TurboLevel 1· GAIA Leaderboard data15.1%
209Gemini 2.5 Flash· GAIA Leaderboard data14.6%
210GPT 5.2Level 3· GAIA Leaderboard data14.3%
211Qwen3.6 35B A3BLevel 3· GAIA Leaderboard data14.3%
212Gemini 2.0 FlashLevel 1· GAIA Leaderboard data14%
213Meta Llama 3 70B Instruct Turbo· GAIA Leaderboard data14%
214Qwen3.5 9B· GAIA Leaderboard data14%
215Qwen3 Next 80B A3B Instruct· GAIA Leaderboard data13%
216qwen2p5-coder-14bLevel 1· GAIA Leaderboard data12.9%
217Mistral SmallLevel 2· GAIA Leaderboard data12.6%
218DeepSeek ChatLevel 3· GAIA Leaderboard data12.2%
219Qwen3.5 9BLevel 2· GAIA Leaderboard data11.9%
220Qwen3.6 35B A3BLevel 2· GAIA Leaderboard data11.9%
221Llama 3 70B InstructLevel 2· GAIA Leaderboard data11.6%
222Gemini 2.5 FlashLevel 2· GAIA Leaderboard data10.7%
223Qwen 3 Next 80bLevel 2· GAIA Leaderboard data10.7%
224Qwen3 Next 80B A3B InstructLevel 2· GAIA Leaderboard data10.7%
225Gemini 3.5 FlashLevel 3· GAIA Leaderboard data10.2%
226mistralLevel 3· GAIA Leaderboard data10.2%
227Mistral SmallLevel 3· GAIA Leaderboard data10.2%
228Qwen 3 Next 80bLevel 3· GAIA Leaderboard data10.2%
229Qwen3 Next 80B A3B InstructLevel 3, Vertex MaaS· GAIA Leaderboard data10.2%
230Claude Sonnet 4.5Level 2· GAIA Leaderboard data10.1%
231qwen2p5-coder-14b· GAIA Leaderboard data9.3%
232qwen2p5-coder-14bLevel 2· GAIA Leaderboard data8.8%
233GPT 5.1Level 3· GAIA Leaderboard data8.2%
234GPT-4o miniLevel 3· GAIA Leaderboard data8.2%
235Qwen3 Next 80B A3B InstructLevel 3, Vertex AI MaaS· GAIA Leaderboard data8.2%
236Llama 3 70B InstructLevel 3· GAIA Leaderboard data7.7%
237Qwen2.5 7B InstructLevel 1· GAIA Leaderboard data6.5%
238Gemini 2.0 Flash· GAIA Leaderboard data6.3%
239Claude Sonnet 4.5Level 3· GAIA Leaderboard data6.1%
240Meta Llama 3 70B Instruct TurboLevel 3· GAIA Leaderboard data6.1%
241Qwen3.5 9BLevel 3· GAIA Leaderboard data6.1%
242GPT 4· GAIA Leaderboard data6.1%
243DeepSeek R1 Distill Qwen 7BLevel 1· GAIA Leaderboard data5.4%
244GPT 5Level 1· GAIA Leaderboard data5.4%
245Qwen2.5 7B Instruct· GAIA Leaderboard data5.3%
246Qwen2.5 7B InstructLevel 2· GAIA Leaderboard data5%
247Gemini 2.5 FlashLevel 3· GAIA Leaderboard data4.1%
248GPT 5Level 3· GAIA Leaderboard data4.1%
249Qwen2.5 7B InstructLevel 3· GAIA Leaderboard data4.1%
250qwen2p5-coder-14bLevel 3· GAIA Leaderboard data4.1%

Showing 201–250 of 291 results.

Data credits: Steel.dev Agent Leaderboards, MIT (licence) · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.