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
301Mixtral 8x7B Instruct v0.1· Hugging Face Open LLM Leaderboard data7%
302Qwen1.5-7B-Chat· Hugging Face Open LLM Leaderboard data7%
303Nous-Hermes-2-Mixtral-8x7B-SFT· Hugging Face Open LLM Leaderboard data6.9%
304L3 8B Lunaris V1v1· Hugging Face Open LLM Leaderboard data6.8%
305Granite 3.0 2B Instruct· Hugging Face Open LLM Leaderboard data6.6%
306Qwen1.5-7B· Hugging Face Open LLM Leaderboard data6.5%
307Qwen1.5-1.8B-Chat· Hugging Face Open LLM Leaderboard data6.4%
308Llama 3 70b Instruct Abliterated v3.5· Hugging Face Open LLM Leaderboard data6.3%
309Qwen1.5-14B· Hugging Face Open LLM Leaderboard data5.9%
310Gemma 1.1 7B IT· Hugging Face Open LLM Leaderboard data5.8%
311mistral-nemo-base-2407· Hugging Face Open LLM Leaderboard data5.8%
312NousResearch: Hermes 2 Pro - Llama-3 8B· Hugging Face Open LLM Leaderboard data5.7%
313qwen2p5-coder-14b· Hugging Face Open LLM Leaderboard data5.7%
314fimbulvetr-11b-v2· Hugging Face Open LLM Leaderboard data5.6%
315Mistral 7B v0.1· Hugging Face Open LLM Leaderboard data5.6%
316mistral-7b-v0.3· Hugging Face Open LLM Leaderboard data5.6%
317Qwen2.5 Coder 7B Instruct· Hugging Face Open LLM Leaderboard data5.6%
318Mistral Nemo· Hugging Face Open LLM Leaderboard data5.4%
319Zephyr 7B Beta· Hugging Face Open LLM Leaderboard data5.4%
320Granite 3.1 2B Instruct· Hugging Face Open LLM Leaderboard data5.3%
321gemma-7b· Hugging Face Open LLM Leaderboard data4.9%
322Llama 3 70B Instruct· Hugging Face Open LLM Leaderboard data4.9%
323Qwen2.5 1.5B· Hugging Face Open LLM Leaderboard data4.7%
324Gemma 7B IT· Hugging Face Open LLM Leaderboard data4.6%
325OpenHermes 2.5 Mistral 7B· Hugging Face Open LLM Leaderboard data4.5%
326OpenHermes-2-Mistral-7B· Hugging Face Open LLM Leaderboard data4.5%
327Yi-6B200k· Hugging Face Open LLM Leaderboard data4.3%
328granite-3.0-3b-a800m-instruct· Hugging Face Open LLM Leaderboard data4.1%
329Mistral (7B) Instruct v0.3· Hugging Face Open LLM Leaderboard data3.9%
330Gemma 2B IT· Hugging Face Open LLM Leaderboard data3.8%
331Qwen1.5-4B· Hugging Face Open LLM Leaderboard data3.6%
332Mistral 7B Instruct v0.2· Hugging Face Open LLM Leaderboard data3.5%
333Llama 3.2 1B Instruct· Hugging Face Open LLM Leaderboard data3.4%
334Gemma 2 2B IT· Hugging Face Open LLM Leaderboard data3.2%
335Starcoder2 15b· Hugging Face Open LLM Leaderboard data3.1%
336Qwen2.5 3B Instruct· Hugging Face Open LLM Leaderboard data3%
337Mistral-7B-OpenOrca· Hugging Face Open LLM Leaderboard data2.9%
338phi-2· Hugging Face Open LLM Leaderboard data2.9%
339Qwen1.5-14B-Chat· Hugging Face Open LLM Leaderboard data2.7%
340Gemma 1.1 2B IT· Hugging Face Open LLM Leaderboard data2.6%
341Qwen1.5-0.5B-Chat· Hugging Face Open LLM Leaderboard data2.6%
342Yi-6B· Hugging Face Open LLM Leaderboard data2.6%
343Qwen1.5-4B-Chat· Hugging Face Open LLM Leaderboard data2.2%
344DeepSeek R1 Distill Llama 70B· Hugging Face Open LLM Leaderboard data2%
345Qwen2.5 Math 1.5b Instruct· Hugging Face Open LLM Leaderboard data2%
346Llama-2-70b-chat-hf· Hugging Face Open LLM Leaderboard data1.9%
347Qwen 2 (1.5B)· Hugging Face Open LLM Leaderboard data1.9%
348Nous-Hermes-llama-2-7b· Hugging Face Open LLM Leaderboard data1.8%
349gemma2-2b· Hugging Face Open LLM Leaderboard data1.7%
350Qwen2 1.5b Instruct· Hugging Face Open LLM Leaderboard data1.6%

Showing 301–350 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.