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
101GPT-4 TurboLevel 1· GAIA Leaderboard data50.9%
102OpenAI o3Level 1· GAIA Leaderboard data50.5%
103Qwen3 Next 80B A3B InstructLevel 1, Vertex MaaS· GAIA Leaderboard data50.5%
104DeepSeek V4 FlashLevel 2· GAIA Leaderboard data50.3%
105Claude Sonnet 4Level 2· GAIA Leaderboard data49.7%
106Claude Sonnet 4.6Level 2· GAIA Leaderboard data49.7%
107Gemini 3.1 Pro Previewopenrouter· GAIA Leaderboard data47.2%
108GPT 4.1· GAIA Leaderboard data47.2%
109GPT 4.1Level 2· GAIA Leaderboard data47.2%
110DeepSeek V3.1Level 1· GAIA Leaderboard data45.2%
111GPT-5.6 TerraLevel 3· GAIA Leaderboard data44.9%
112GPT 4o· GAIA Leaderboard data44.2%
113Gemini 3.1 Pro PreviewLevel 2, openrouter· GAIA Leaderboard data44%
114GLM 5.3 Flash· GAIA Leaderboard data43.9%
115GPT 4oLevel 2· GAIA Leaderboard data43%
116Qwen3 Next 80B A3B InstructLevel 1, Vertex AI MaaS· GAIA Leaderboard data43%
117GLM 5.1· GAIA Leaderboard data42.5%
118o4 Mini· GAIA Leaderboard data42.5%
119Gemini 3 Flash· GAIA Leaderboard data41.5%
120DeepSeek V4.1 Flash· GAIA Leaderboard data41.2%
121GPT-5.4 miniLevel 3· GAIA Leaderboard data40.8%
122GLM 5.3 FlashLevel 2· GAIA Leaderboard data40.3%
123DeepSeek V4.1 FlashLevel 2· GAIA Leaderboard data39%
124DeepSeek FlashLevel 3· GAIA Leaderboard data38.8%
125GPT 4ogpt-4o-2024-08-06· GAIA Leaderboard data38.2%
126GLM 5.1Level 2· GAIA Leaderboard data37.7%
127GPT-4 Turbo· GAIA Leaderboard data37.6%
128DeepSeek V3.2· GAIA Leaderboard data37.2%
129OpenAI o3· GAIA Leaderboard data37.2%
130Gemini 3 FlashLevel 2· GAIA Leaderboard data37.1%
131o4 MiniLevel 2· GAIA Leaderboard data37.1%
132Claude Sonnet 4.6Level 3· GAIA Leaderboard data36.7%
133GPT-5.6 LunaLevel 3· GAIA Leaderboard data36.7%
134Claude Haiku 4.5Level 1· GAIA Leaderboard data36.6%
135Gemini 3.5 FlashLevel 1· GAIA Leaderboard data36.6%
136GPT 4oLevel 2, gpt-4o-2024-08-06· GAIA Leaderboard data36.5%
137GPT-4 TurboLevel 2· GAIA Leaderboard data36%
138QwQ 32BLevel 1· GAIA Leaderboard data35.8%
139mistralLevel 1· GAIA Leaderboard data35.5%
140Qwen 3 Next 80bLevel 1· GAIA Leaderboard data35.5%
141o1Level 3· GAIA Leaderboard data34.6%
142DeepSeek ChatLevel 1· GAIA Leaderboard data34.4%
143GPT 5.1Level 1· GAIA Leaderboard data33.3%
144DeepSeek V3.2Level 2· GAIA Leaderboard data32.7%
145DeepSeek V4 FlashLevel 3· GAIA Leaderboard data32.7%
146GPT 5.2Level 1· GAIA Leaderboard data32.3%
147GPT-4o miniLevel 1· GAIA Leaderboard data32.3%
148claude· GAIA Leaderboard data32.2%
149OpenAI o3Level 2· GAIA Leaderboard data32.1%
150Claude Sonnet 4.5Level 1· GAIA Leaderboard data31.2%

Showing 101–150 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.