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
151Claude Sonnet 3.7Level 3· GAIA Leaderboard data30.8%
152Llama 3 70B InstructLevel 1· GAIA Leaderboard data30.2%
153Mistral SmallLevel 1· GAIA Leaderboard data30.1%
154Qwen3 32BLevel 1· GAIA Leaderboard data30.1%
155Qwen3 Next 80B A3B InstructVertex MaaS· GAIA Leaderboard data29.6%
156OpenAI o3Level 3· GAIA Leaderboard data28.6%
157DeepSeek V3.1· GAIA Leaderboard data27.9%
158Claude 3.5 SonnetLevel 3· GAIA Leaderboard data26.9%
159Gemini 2.5 FlashLevel 1· GAIA Leaderboard data26.9%
160Qwen3.6 35B A3BLevel 1· GAIA Leaderboard data26.9%
161DeepSeek V4.1 FlashLevel 3· GAIA Leaderboard data26.5%
162Gemini 3 FlashLevel 3· GAIA Leaderboard data26.5%
163Gemini 3.1 Pro PreviewLevel 3, openrouter· GAIA Leaderboard data26.5%
164Qwen3 Next 80B A3B InstructVertex AI MaaS· GAIA Leaderboard data26.2%
165DeepSeek V3.1Level 2· GAIA Leaderboard data25.8%
166GPT 5.2· GAIA Leaderboard data25.2%
167Gemini 3.5 Flash· GAIA Leaderboard data24.9%
168GPT 5.2Level 2· GAIA Leaderboard data24.5%
169qwen-coderLevel 1· GAIA Leaderboard data24.5%
170Claude Sonnet 4Level 3· GAIA Leaderboard data24.5%
171o4 MiniLevel 3· GAIA Leaderboard data24.5%
172Claude Haiku 4.5· GAIA Leaderboard data23.6%
173DeepSeek Chat· GAIA Leaderboard data23.6%
174Claude Haiku 4.5Level 2· GAIA Leaderboard data23.3%
175Qwen3 Next 80B A3B InstructLevel 2, Vertex MaaS· GAIA Leaderboard data23.3%
176mistral· GAIA Leaderboard data22.9%
177Gemini 3.5 FlashLevel 2· GAIA Leaderboard data22.6%
178GPT-4o mini· GAIA Leaderboard data22.6%
179Qwen3 32BLevel 2· GAIA Leaderboard data22%
180Qwen3 Next 80B A3B InstructLevel 2, Vertex AI MaaS· GAIA Leaderboard data22%
181Qwen3 32B· GAIA Leaderboard data21.6%
182Qwen3 Next 80B A3B InstructLevel 1· GAIA Leaderboard data21.5%
183Qwen3.5 9BLevel 1· GAIA Leaderboard data21.5%
184GPT-4o miniLevel 2· GAIA Leaderboard data21.4%
185QwQ 32B· GAIA Leaderboard data21.2%
186DeepSeek ChatLevel 2· GAIA Leaderboard data20.8%
187GPT 5.1· GAIA Leaderboard data19.6%
188mistralLevel 2· GAIA Leaderboard data19.5%
189GPT 4oLevel 3· GAIA Leaderboard data19.2%
190GLM 5.1Level 3· GAIA Leaderboard data18.4%
191GPT 4.1Level 3· GAIA Leaderboard data18.4%
192GPT 4oLevel 3, gpt-4o-2024-08-06· GAIA Leaderboard data18.4%
193Qwen 3 Next 80b· GAIA Leaderboard data18.3%
194Mistral Small· GAIA Leaderboard data17.6%
195QwQ 32BLevel 2· GAIA Leaderboard data17.4%
196Step 3.5 FlashStep 3.5 technical report · Table 11 · 256K · no-tools baseline· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.17%
197Llama 3 70B Instruct· GAIA Leaderboard data17%
198qwen-coder· GAIA Leaderboard data17%
199Qwen3.6 35B A3B· GAIA Leaderboard data16.9%
200DeepSeek V3.2Level 3· GAIA Leaderboard data16.3%

Showing 151–200 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.