Skip to content

Benchmark

Dtbench Accuracy

Scores are percentages. Higher is better.

Models measured
170
Scores recorded
239
Sources
1
IQ weight
1.22

Score distribution

79.8%
Mean
84.5%
Median
16.6%
Std dev
52.5%
10th percentile
97.1%
90th percentile
41.7%
Min
41.7%99.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.22 for chat models, measured over 170 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
151Qwen3 235B A22B Instruct 2507· Epoch AI Benchmarking Hub data78.4%
152Qwen3.6 27B· Epoch AI Benchmarking Hub data78.1%
153o4 Minihigh· Epoch AI Benchmarking Hub data77.6%
154Claude Sonnet 4· Epoch AI Benchmarking Hub data77.1%
155Qwen3.6 Flash· Epoch AI Benchmarking Hub data77.1%
156Gemini 3.1 Flash Litehigh· Epoch AI Benchmarking Hub data76.8%
157Gemini 2.5 Flash· Epoch AI Benchmarking Hub data76.5%
158GPT OSS 120Bhigh· Epoch AI Benchmarking Hub data76.3%
159Qwen 3 235B A22B· Epoch AI Benchmarking Hub data75.7%
160Mistral Medium2604· Epoch AI Benchmarking Hub data75.5%
161Claude Haiku 5.5none· Epoch AI Benchmarking Hub data75.2%
162Gemma 4 26B A4B· Epoch AI Benchmarking Hub data74.9%
163o1high· Epoch AI Benchmarking Hub data74.7%
164Qwen3.6 35B A3B· Epoch AI Benchmarking Hub data73.9%
165Claude Haiku 4.5· Epoch AI Benchmarking Hub data73.6%
166Claude Sonnet 5low· Epoch AI Benchmarking Hub data73.6%
167Qwen3.5 9B· Epoch AI Benchmarking Hub data71.2%
168Mistral Small2603· Epoch AI Benchmarking Hub data70.9%
169Qwen3 30B A3B2507, thinking· Epoch AI Benchmarking Hub data69.3%
170GPT-5.6 Lunanone· Epoch AI Benchmarking Hub data69.3%
171GPT-4.1 mini· Epoch AI Benchmarking Hub data68.8%
172o3 Minihigh· Epoch AI Benchmarking Hub data68.8%
173GPT 4.1· Epoch AI Benchmarking Hub data68.3%
174GPT OSS 20bhigh· Epoch AI Benchmarking Hub data68%
175Claude 3.5 Sonnet· Epoch AI Benchmarking Hub data67.8%
176Mistral Medium 3.1· Epoch AI Benchmarking Hub data67.5%
177Qwen3 32B· Epoch AI Benchmarking Hub data67.5%
178Qwen3 30B A3B Instruct· Epoch AI Benchmarking Hub data67.2%
179Grok 2· Epoch AI Benchmarking Hub data65.2%
180Mistral Large 3· Epoch AI Benchmarking Hub data65.1%
181DeepSeek V3· Epoch AI Benchmarking Hub data64.8%
182GPT 4o· Epoch AI Benchmarking Hub data64.5%
183Qwen3 14B· Epoch AI Benchmarking Hub data64%
184Gemini 2.0 Flash· Epoch AI Benchmarking Hub data63.2%
185Qwen2.5 72B Instruct· Epoch AI Benchmarking Hub data62.9%
186Gemini 2.5 Flash-Litethinking· Epoch AI Benchmarking Hub data62.8%
187DeepSeek Chat· Epoch AI Benchmarking Hub data62.7%
188GPT 4· Epoch AI Benchmarking Hub data62.7%
189GPT-5 nanohigh· Epoch AI Benchmarking Hub data62.7%
190Mistral Medium 3· Epoch AI Benchmarking Hub data62.3%
191Llama 4 Maverick· Epoch AI Benchmarking Hub data61.9%
192Claude 3 Opus· Epoch AI Benchmarking Hub data61.6%
193GPT-4 Turbo· Epoch AI Benchmarking Hub data61.6%
194Llama 3.1 405B Instruct· Epoch AI Benchmarking Hub data61.4%
195C4AI Command A 03 2025· Epoch AI Benchmarking Hub data61.3%
196Magistral Small 2509· Epoch AI Benchmarking Hub data61.3%
197Mistral Large2407· Epoch AI Benchmarking Hub data61.2%
198Mistral Large2411· Epoch AI Benchmarking Hub data60.8%
199Qwen3 30B A3B· Epoch AI Benchmarking Hub data60.3%
200Llama 3.1 70B Instruct· Epoch AI Benchmarking Hub data60%

Showing 151–200 of 239 results.

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