Skip to content

Benchmark

GPQA Diamond

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

Models and systems measured
282
Scores recorded
1623
Sources
4
IQ weight
1.72

Score distribution

79.3%
Mean
84.5%
Median
15.7%
Std dev
56.8%
10th percentile
92.2%
90th percentile
0.168%
Min
0.168%96.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.72 for chat models, measured over 272 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 system
1201o1medium· Epoch AI Benchmarking Hub data75.8%
1202Qwen 3.8 27BOpenRouter harness, qwen/qwen3.8-27b-20260814, Phala, public 90-day aggregate· OpenRouter benchmark exports75.7%
1203MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports75.7%
1204Qwen3 Coder NextOpenRouter harness, qwen/qwen3-coder-next-2025-02-03, StreamLake, public 90-day aggregate· OpenRouter benchmark exports75.6%
1205MiMo V2.6 FlashOpenRouter harness, xiaomi/mimo-v2.6-flash-20260921, Novita, public 90-day aggregate· OpenRouter benchmark exports75.6%
1206Mistral SmallOpenRouter harness, mistralai/mistral-small-2603, Venice, public 90-day aggregate· OpenRouter benchmark exports75.6%
1207Nemotron 3.5 Lightning 30B A3Bno tools· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75.4%
1208Gemma 4 26B A4BOpenRouter harness, google/gemma-4-26b-a4b-it-20260403, NextBit, public 90-day aggregate· OpenRouter benchmark exports75.4%
1209Claude Sonnet 4extended thinking · up to 64K thinking tokens· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75.4%
1210DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports75.4%
1211Mistral SmallOpenRouter harness, mistralai/mistral-small-2603, Mistral, public 90-day aggregate· OpenRouter benchmark exports75.3%
1212GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, Io Net, public 90-day aggregate· OpenRouter benchmark exports75.3%
1213o4 Minilow· Epoch AI Benchmarking Hub data75.3%
1214Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, PrimeIntellect, public 90-day aggregate· OpenRouter benchmark exports75.3%
1215GLM 4.7 Flash· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.
Values from 2 sources
75.2%
1216MiMo V2.6 FlashOpenRouter harness, xiaomi/mimo-v2.6-flash-20260921, Io Net, public 90-day aggregate· OpenRouter benchmark exports75.1%
1217Gemma 4 26B A4BOpenRouter harness, google/gemma-4-26b-a4b-it-20260403, Makora, public 90-day aggregate· OpenRouter benchmark exports75.1%
1218Qwen3 Coder NextOpenRouter harness, qwen/qwen3-coder-next-2025-02-03, auto-routing, public 90-day aggregate· OpenRouter benchmark exports75.1%
1219GPT-5 minihigh· Epoch AI Benchmarking Hub data75%
1220Qwen3 Coder NextOpenRouter harness, qwen/qwen3-coder-next-2025-02-03, Alibaba, public 90-day aggregate· OpenRouter benchmark exports74.9%
1221Claude Opus 4no extended thinking· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.74.9%
1222Ling 3.0 FlashOpenRouter harness, inclusionai/ling-3.0-flash-20260723, auto-routing, public 90-day aggregate· OpenRouter benchmark exports74.9%
1223MiMo V2.6 FlashOpenRouter harness, xiaomi/mimo-v2.6-flash-20260921, Darkbloom, public 90-day aggregate· OpenRouter benchmark exports74.8%
1224Nemotron 3.5 LightningOpenRouter harness, nvidia/nemotron-3.5-lightning-20260807, Wafer, public 90-day aggregate· OpenRouter benchmark exports74.8%
1225GLM 4.6OpenRouter harness, z-ai/glm-4.6, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports74.7%
1226GPT 5.4none· Epoch AI Benchmarking Hub data74.7%
1227Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, Alibaba, public 90-day aggregate· OpenRouter benchmark exports74.7%
1228Grok 3 Minihigh, beta· Epoch AI Benchmarking Hub data74.6%
1229Nemotron 3.5 LightningOpenRouter harness, nvidia/nemotron-3.5-lightning-20260807, ModelRun, public 90-day aggregate· OpenRouter benchmark exports74.6%
1230Qwen3 Coder NextOpenRouter harness, qwen/qwen3-coder-next-2025-02-03, Parasail, public 90-day aggregate· OpenRouter benchmark exports74.5%
1231GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, auto-routing, public 90-day aggregate· OpenRouter benchmark exports74.5%
1232Gemma 4 26B A4BOpenRouter harness, google/gemma-4-26b-a4b-it-20260403, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports74.4%
1233o3 Minimedium· Epoch AI Benchmarking Hub data74.3%
1234Gemini 3.1 Flash Litelow· Epoch AI Benchmarking Hub data74.2%
1235Gemini 3.5 Flash Liteminimal· Epoch AI Benchmarking Hub data74.2%
1236o1low· Epoch AI Benchmarking Hub data74.2%
1237GPT OSS 20bofficial model card Table 3 · reasoning high · with tools; Table 3 does not name the exact tool · pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.74.2%
1238GLM 5.1OpenRouter harness, z-ai/glm-5.1-20260406, Baidu, public 90-day aggregate· OpenRouter benchmark exports74.2%
1239GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, Novita, public 90-day aggregate· OpenRouter benchmark exports74.1%
1240Claude Sonnet 4.6OpenRouter harness, anthropic/claude-4.6-sonnet-20260217, Azure, public 90-day aggregate· OpenRouter benchmark exports74.1%
1241Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, StreamLake, public 90-day aggregate· OpenRouter benchmark exports74.1%
1242Qwen3.5 9BOpenRouter harness, qwen/qwen3.5-9b-20260310, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports74.1%
1243Claude Sonnet 4.6OpenRouter harness, anthropic/claude-4.6-sonnet-20260217, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports74%
1244Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, Nebius, public 90-day aggregate· OpenRouter benchmark exports74%
1245Nemotron 3.5 LightningOpenRouter harness, nvidia/nemotron-3.5-lightning-20260807, Io Net, public 90-day aggregate· OpenRouter benchmark exports74%
1246Qwen3 Coder NextOpenRouter harness, qwen/qwen3-coder-next-2025-02-03, Novita, public 90-day aggregate· OpenRouter benchmark exports73.9%
1247GLM 5OpenRouter harness, z-ai/glm-5-20260211, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports73.9%
1248Gemini 2.5 FlashOpenRouter harness, google/gemini-2.5-flash, Google, public 90-day aggregate· OpenRouter benchmark exports73.8%
1249GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports73.7%
1250Claude Sonnet 4.6OpenRouter harness, anthropic/claude-4.6-sonnet-20260217, auto-routing, public 90-day aggregate· OpenRouter benchmark exports73.7%

Showing 1,201–1,250 of 1,621 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 system
MiMo V2 Flash Baseofficial Base table · 5-shot· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.55.1%
Qwen3.8-2.4T-A95Blifecycle-page core-benchmark claim· Reported by Alibaba Cloud Model Studio· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.6%

Tokens the model generated per task, as measured by the source. Comparable within this benchmark only.

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (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. · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence) · OpenEvals/leaderboard-data, MIT (licence)