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
1001GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Morph, public 90-day aggregate· OpenRouter benchmark exports82.1%
1002GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Makora, public 90-day aggregate· OpenRouter benchmark exports82.1%
1003DeepSeek V3.2OpenRouter harness, deepseek/deepseek-v3.2-20251201, Venice, public 90-day aggregate· OpenRouter benchmark exports82.1%
1004GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Io Net, public 90-day aggregate· OpenRouter benchmark exports82.1%
1005MiMo V2.5 ProOpenRouter harness, xiaomi/mimo-v2.5-pro-20260422, auto-routing, public 90-day aggregate· OpenRouter benchmark exports82.1%
1006Gemini 3.1 Flash LiteOpenRouter harness, google/gemini-3.1-flash-lite-20260507, auto-routing, public 90-day aggregate· OpenRouter benchmark exports82.1%
1007Qwen 3.8 27BOpenRouter harness, qwen/qwen3.8-27b-20260814, Mancer 2, public 90-day aggregate· OpenRouter benchmark exports82.1%
1008Qwen3.5 35B A3BOpenRouter harness, qwen/qwen3.5-35b-a3b-20260224, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports82.1%
1009GLM 5SenseNova 6.7 Flash-Lite official benchmark chart · chart setting· Reported by SenseTime / OpenSenseNova· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82%
1010Muse Glimmer 30BOpenRouter harness, meta/muse-glimmer-30b-20260810, Together, public 90-day aggregate· OpenRouter benchmark exports82%
1011GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, StreamLake, public 90-day aggregate· OpenRouter benchmark exports82%
1012GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Baidu, public 90-day aggregate· OpenRouter benchmark exports82%
1013Qwen 3.8 27BOpenRouter harness, qwen/qwen3.8-27b-20260814, Darkbloom, public 90-day aggregate· OpenRouter benchmark exports82%
1014Muse Glimmer 30BOpenRouter harness, meta/muse-glimmer-30b-20260810, Fireworks, public 90-day aggregate· OpenRouter benchmark exports81.9%
1015Qwen 3.8 27BOpenRouter harness, qwen/qwen3.8-27b-20260814, Tenstorrent, public 90-day aggregate· OpenRouter benchmark exports81.9%
1016GLM 5OpenRouter harness, z-ai/glm-5-20260211, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports81.9%
1017Inkling FP4OpenRouter harness, thinkingmachines/inkling-20260715, BaseTen, public 90-day aggregate· OpenRouter benchmark exports81.9%
1018Kimi K2.7 CodeOpenRouter harness, moonshotai/kimi-k2.7-code-20260612, Alibaba, public 90-day aggregate· OpenRouter benchmark exports81.9%
1019DeepSeek V3.2OpenRouter harness, deepseek/deepseek-v3.2-20251201, Phala, public 90-day aggregate· OpenRouter benchmark exports81.8%
1020Gemini 3.1 Flash Litehigh· Epoch AI Benchmarking Hub data81.8%
1021OpenAI o3high· Epoch AI Benchmarking Hub data81.8%
1022Qwen3.7 Plusnone· Epoch AI Benchmarking Hub data81.8%
1023GLM 5OpenRouter harness, z-ai/glm-5-20260211, StreamLake, public 90-day aggregate· OpenRouter benchmark exports81.8%
1024GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Tenstorrent, public 90-day aggregate· OpenRouter benchmark exports81.8%
1025Qwen 3.8 27BOpenRouter harness, qwen/qwen3.8-27b-20260814, Near AI, public 90-day aggregate· OpenRouter benchmark exports81.8%
1026MiniMax M3OpenRouter harness, minimax/minimax-m3-20260531, SambaNova, public 90-day aggregate· OpenRouter benchmark exports81.8%
1027MiMo V2.5OpenRouter harness, xiaomi/mimo-v2.5-20260422, StreamLake, public 90-day aggregate· OpenRouter benchmark exports81.8%
1028Qwen3.5 35B A3BOpenRouter harness, qwen/qwen3.5-35b-a3b-20260224, Parasail, public 90-day aggregate· OpenRouter benchmark exports81.7%
1029Claude Sonnet 4.532k· Epoch AI Benchmarking Hub data81.7%
1030Qwen3.6 27BOpenRouter harness, qwen/qwen3.6-27b-20260422, Reka, public 90-day aggregate· OpenRouter benchmark exports81.7%
1031GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, OpenInference, public 90-day aggregate· OpenRouter benchmark exports81.7%
1032DeepSeek V3.2OpenRouter harness, deepseek/deepseek-v3.2-20251201, Friendli, public 90-day aggregate· OpenRouter benchmark exports81.7%
1033Qwen3.5 9B· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.
Values from 2 sources
81.7%
1034GLM 5.1OpenRouter harness, z-ai/glm-5.1-20260406, Novita, public 90-day aggregate· OpenRouter benchmark exports81.7%
1035Qwen3.5 35B A3BOpenRouter harness, qwen/qwen3.5-35b-a3b-20260224, Venice, public 90-day aggregate· OpenRouter benchmark exports81.6%
1036GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, KT, public 90-day aggregate· OpenRouter benchmark exports81.6%
1037Claude Sonnet 4.5OpenRouter harness, anthropic/claude-4.5-sonnet-20250929, Anthropic, public 90-day aggregate· OpenRouter benchmark exports81.6%
1038Claude Sonnet 4.5OpenRouter harness, anthropic/claude-4.5-sonnet-20250929, auto-routing, public 90-day aggregate· OpenRouter benchmark exports81.6%
1039Claude Sonnet 4.5OpenRouter harness, anthropic/claude-4.5-sonnet-20250929, Azure, public 90-day aggregate· OpenRouter benchmark exports81.5%
1040Qwen3.5 35B A3BOpenRouter harness, qwen/qwen3.5-35b-a3b-20260224, auto-routing, public 90-day aggregate· OpenRouter benchmark exports81.5%
1041GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Relace, public 90-day aggregate· OpenRouter benchmark exports81.5%
1042GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, Tenstorrent, public 90-day aggregate· OpenRouter benchmark exports81.5%
1043Kimi K2.5OpenRouter harness, moonshotai/kimi-k2.5-0127, StreamLake, public 90-day aggregate· OpenRouter benchmark exports81.5%
1044GLM 5.1OpenRouter harness, z-ai/glm-5.1-20260406, auto-routing, public 90-day aggregate· OpenRouter benchmark exports81.5%
1045Qwen 3.8 27BOpenRouter harness, qwen/qwen3.8-27b-20260814, Io Net, public 90-day aggregate· OpenRouter benchmark exports81.5%
1046Gemini 3.1 Flash LiteOpenRouter harness, google/gemini-3.1-flash-lite-20260507, Google AI Studio, public 90-day aggregate· OpenRouter benchmark exports81.4%
1047Claude Sonnet 4.5OpenRouter harness, anthropic/claude-4.5-sonnet-20250929, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports81.4%
1048DeepSeek V3.2OpenRouter harness, deepseek/deepseek-v3.2-20251201, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports81.4%
1049GLM 5OpenRouter harness, z-ai/glm-5-20260211, Venice, public 90-day aggregate· OpenRouter benchmark exports81.3%
1050Claude Sonnet 4.5OpenRouter harness, anthropic/claude-4.5-sonnet-20250929, Google, public 90-day aggregate· OpenRouter benchmark exports81.3%

Showing 1,001–1,050 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)