Skip to content

Benchmark

WeirdML

Scores are percentages.

Models measured
129
Scores recorded
164
Sources
1
IQ weight
2.56

Score distribution

49.8%
Mean
46.1%
Median
21.7%
Std dev
22.2%
10th percentile
82.9%
90th percentile
1.7%
Min
1.7%93.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.56 for chat models, measured over 129 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 OSS 120Bmedium· Epoch AI Benchmarking Hub data41.9%
102DeepSeek R1· Epoch AI Benchmarking Hub data41.6%
103Qwen3 Coder· Epoch AI Benchmarking Hub data41.2%
104Qwen 3 235B A22B Thinking 2507· Epoch AI Benchmarking Hub data41%
105Gemini 2.5 Flash Preview 05-20 (no think)16k· Epoch AI Benchmarking Hub data41%
106gemini-2.5-flash-preview-04-1716k· Epoch AI Benchmarking Hub data41%
107GPT OSS 20bhigh· Epoch AI Benchmarking Hub data40.9%
108GLM 4.5thinking· Epoch AI Benchmarking Hub data40.6%
109Claude 3.5 Sonnet· Epoch AI Benchmarking Hub data40%
110GPT-5 Chat· Epoch AI Benchmarking Hub data39.8%
111Qwen3.5 27B· Epoch AI Benchmarking Hub data39.5%
112DeepSeek V3.2 Expthinking· Epoch AI Benchmarking Hub data39.5%
113GPT-4.5 Preview· Epoch AI Benchmarking Hub data39.4%
114Kimi K2 Instruct· Epoch AI Benchmarking Hub data39.4%
115kimi-k2-0711-preview· Epoch AI Benchmarking Hub data39.4%
116GPT 4.1· Epoch AI Benchmarking Hub data39%
117Gemini 3.5 Flash Litehigh· Epoch AI Benchmarking Hub data39%
118Qwen3 235B A22B Instruct 2507· Epoch AI Benchmarking Hub data38.7%
119DeepSeek V3.1thinking· Epoch AI Benchmarking Hub data38.4%
120GPT-5 nanohigh· Epoch AI Benchmarking Hub data38.1%
121Nemotron 3 Super· Epoch AI Benchmarking Hub data38%
122GPT 5.4 Nanonone· Epoch AI Benchmarking Hub data38%
123GPT-5.4 mininone· Epoch AI Benchmarking Hub data37.9%
124GPT-4.1 mini· Epoch AI Benchmarking Hub data37.6%
125DeepSeek V3.1· Epoch AI Benchmarking Hub data37.5%
126Qwen 3 235B A22B· Epoch AI Benchmarking Hub data37.3%
127Grok 3· Epoch AI Benchmarking Hub data37.2%
128MiniMax M2.7· Epoch AI Benchmarking Hub data37%
129GPT OSS 20bmedium· Epoch AI Benchmarking Hub data36.8%
130O1 Minimedium· Epoch AI Benchmarking Hub data36.3%
131DeepSeek V3· Epoch AI Benchmarking Hub data36.1%
132Gemini 2.5 Flash-Lite16k· Epoch AI Benchmarking Hub data35.2%
133Gemini 2.5 Flash-Litethinking· Epoch AI Benchmarking Hub data35.2%
134Gemma 4 26B A4B· Epoch AI Benchmarking Hub data35.2%
135Grok Code Fast 1· Epoch AI Benchmarking Hub data35.1%
136Qwen3.6 35B A3B· Epoch AI Benchmarking Hub data34.5%
137Qwen3 Coder Next· Epoch AI Benchmarking Hub data34.4%
138Mistral Medium 3.1· Epoch AI Benchmarking Hub data33.1%
139Inkling FP4high· Epoch AI Benchmarking Hub data32.3%
140Claude Haiku 3.5· Epoch AI Benchmarking Hub data30.7%
141Qwen3 30B A3B· Epoch AI Benchmarking Hub data29.8%
142GPT-5 nanolow· Epoch AI Benchmarking Hub data25.9%
143Gemini 2.0 Flash· Epoch AI Benchmarking Hub data25.8%
144GPT 4o· Epoch AI Benchmarking Hub data25.1%
145gemini-1.5-flash-002· Epoch AI Benchmarking Hub data24.9%
146Llama 4 Maverick· Epoch AI Benchmarking Hub data24.5%
147Grok 2· Epoch AI Benchmarking Hub data22.2%
148gemini-1.5-pro-002· Epoch AI Benchmarking Hub data22.2%
149Llama 3.1 405B Instruct· Epoch AI Benchmarking Hub data21.4%
150Claude 3 Opus· Epoch AI Benchmarking Hub data19.2%

Showing 101–150 of 164 results.

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence)