Skip to content

Benchmark

GPQA

Science benchmarks measure how well models answer science questions. Scores are percentages. Higher is better.

Models and systems measured
364
Scores recorded
378
Sources
3
IQ weight
0.75

Score distribution

45.3%
Mean
47.5%
Median
34.4%
Std dev
2.9%
10th percentile
89%
90th percentile
0%
Min
0%97%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.75 for chat models, measured over 359 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
251DeepHermes 3 Mistral 24B Preview· Hugging Face Open LLM Leaderboard data16%
252dolphin-2-9-2-qwen2-72b· Hugging Face Open LLM Leaderboard data16%
253Yi-34B· Hugging Face Open LLM Leaderboard data15.5%
254Yi 1.5 34B Chat· Hugging Face Open LLM Leaderboard data15.3%
255Hermes 3 Llama 3.1 70B· Hugging Face Open LLM Leaderboard data14.9%
256Gemma 2 9B IT· Hugging Face Open LLM Leaderboard data14.8%
257Yi-34B200k· Hugging Face Open LLM Leaderboard data14.2%
258gemma2-27b· Hugging Face Open LLM Leaderboard data13.4%
259Qwen 2.5 Coder 32B Instruct· Hugging Face Open LLM Leaderboard data13.2%
260internlm2-5-7b-chat· Hugging Face Open LLM Leaderboard data13%
261Qwen 2.5 Coder 32B· Hugging Face Open LLM Leaderboard data12.9%
262DBRX-instruct· Hugging Face Open LLM Leaderboard data12.2%
263Qwen1.5-110B-Chat· Hugging Face Open LLM Leaderboard data12.2%
264Qwen3.5 0.8B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim11.9%
265Yi 1.5 34B Chat16k· Hugging Face Open LLM Leaderboard data11.7%
266Yi 34B Chat· Hugging Face Open LLM Leaderboard data11.7%
267Phi-3 Medium 128K Instruct· Hugging Face Open LLM Leaderboard data11.5%
268Phi-3 Medium 4K Instruct· Hugging Face Open LLM Leaderboard data11.5%
269Granite 3.0 8B Instruct· Hugging Face Open LLM Leaderboard data11%
270l3-70b-euryale-v2.1· Hugging Face Open LLM Leaderboard data10.9%
271Qwen2 57B A14B Instruct· Hugging Face Open LLM Leaderboard data10.9%
272Qwen2.5 Math 72B Instruct· Hugging Face Open LLM Leaderboard data10.9%
273gemma-2-9b· Hugging Face Open LLM Leaderboard data10.5%
274Aya Expanse 32B· Hugging Face Open LLM Leaderboard data10.1%
275Qwen2.5 7B· Hugging Face Open LLM Leaderboard data10%
276granite-3-1-8b-base· Hugging Face Open LLM Leaderboard data9.5%
277Nous Hermes 2 Mixtral 8x7B DPO· Hugging Face Open LLM Leaderboard data9.5%
278Phi-3 Mini 4K Instruct· Hugging Face Open LLM Leaderboard data9.3%
279Qwen2 VL 7b Instruct· Hugging Face Open LLM Leaderboard data9.3%
280Phi-3 Mini 128K Instruct· Hugging Face Open LLM Leaderboard data9.1%
281DeepSeek LLM 67B Chat· Hugging Face Open LLM Leaderboard data8.8%
282GLM 4 9B 0414· Hugging Face Open LLM Leaderboard data8.8%
283MN 12b Mag Mell R1· Hugging Face Open LLM Leaderboard data8.8%
284Granite 3.2 8b Instruct· Hugging Face Open LLM Leaderboard data8.7%
285GLM-4-9b-chat· Hugging Face Open LLM Leaderboard data8.5%
286Mixtral 8X7b V0.1· Hugging Face Open LLM Leaderboard data8.5%
287Granite 3.1 8B Instruct· Hugging Face Open LLM Leaderboard data8.3%
288Phi-3 Small 8K Instruct· Hugging Face Open LLM Leaderboard data8.3%
289L3 8B Stheno V3.2· Hugging Face Open LLM Leaderboard data8.1%
290Llama 3.1 8B· Hugging Face Open LLM Leaderboard data8.1%
291C4AI Command R7b 12 2024· Hugging Face Open LLM Leaderboard data7.8%
292SOLAR 10.7B Instruct v1.0· Hugging Face Open LLM Leaderboard data7.8%
293OpenChat 3.5· Hugging Face Open LLM Leaderboard data7.7%
294Neural Daredevil 8B abliterated· Hugging Face Open LLM Leaderboard data7.5%
295Qwen1.5-32B-Chat· Hugging Face Open LLM Leaderboard data7.5%
296Meta Llama 3 8b· Hugging Face Open LLM Leaderboard data7.4%
297Qwen1.5-1.8B· Hugging Face Open LLM Leaderboard data7.4%
298Qwen 2 (7B)· Hugging Face Open LLM Leaderboard data7.3%
299qwen2.5-coder-14b-instruct· Hugging Face Open LLM Leaderboard data7.3%
300Aya Expanse 8B· Hugging Face Open LLM Leaderboard data7%

Showing 251–300 of 374 results.

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
Qwen3.5-397B-A17B-BaseQwen3.5 release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.54.6%
Qwen3.7-Plus-BaseQwen3.8-Flash-Next release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.5%
Qwen3.8-27B-BaseQwen3.8-Flash-Next release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.45%
Qwen3.8-Flash-Next-BaseQwen3.8-Flash-Next release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.4%

Data credits: Data by LLM Stats · 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.