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
501GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports88.4%
502Claude Opus 4.6max· Epoch AI Benchmarking Hub data88.4%
503Claude Opus 4.8low· Epoch AI Benchmarking Hub data88.4%
504Qwen3.6 Plus· Epoch AI Benchmarking Hub data88.4%
505Claude Fable 5OpenRouter harness, anthropic/claude-5-fable-20260609, Azure, public 90-day aggregate· OpenRouter benchmark exports88.4%
506GPT-5.6 TerraOpenRouter harness, openai/gpt-5.6-terra-20260709, auto-routing, public 90-day aggregate· OpenRouter benchmark exports88.4%
507GPT-6 LunaOpenRouter harness, openai/gpt-6-luna-pro-20260922, auto-routing, public 90-day aggregate· OpenRouter benchmark exports88.4%
508MiniMax M3.1 FlashOpenRouter harness, minimax/minimax-m3.1-flash-20261019, StreamLake, public 90-day aggregate· OpenRouter benchmark exports88.4%
509Claude Opus 5OpenRouter harness, anthropic/claude-opus-5-20260723, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports88.4%
510Kimi K2.7 CodeOpenRouter harness, moonshotai/kimi-k2.7-code-20260612, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports88.3%
511DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports88.3%
512DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, DigitalOcean, public 90-day aggregate· OpenRouter benchmark exports88.3%
513MiMo V2.6 ProOpenRouter harness, xiaomi/mimo-v2.6-pro-20260921, auto-routing, public 90-day aggregate· OpenRouter benchmark exports88.3%
514DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, Venice, public 90-day aggregate· OpenRouter benchmark exports88.3%
515Claude Opus 5OpenRouter harness, anthropic/claude-opus-5-20260723, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports88.3%
516MiniMax M3.1 FlashOpenRouter harness, minimax/minimax-m3.1-flash-20261019, Minimax, public 90-day aggregate· OpenRouter benchmark exports88.3%
517Claude Fable 5OpenRouter harness, anthropic/claude-5-fable-20260609, Google, public 90-day aggregate· OpenRouter benchmark exports88.3%
518DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, Reka, public 90-day aggregate· OpenRouter benchmark exports88.3%
519Kimi K2.6OpenRouter harness, moonshotai/kimi-k2.6-20260420, auto-routing, public 90-day aggregate· OpenRouter benchmark exports88.3%
520DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, StreamLake, public 90-day aggregate· OpenRouter benchmark exports88.3%
521Inkling FP4xhigh· Epoch AI Benchmarking Hub data88.3%
522DeepSeek V4 ProOpenRouter harness, deepseek/deepseek-v4-pro-20260423, Parasail, public 90-day aggregate· OpenRouter benchmark exports88.3%
523Qwen3.5 397B A17BOpenRouter harness, qwen/qwen3.5-397b-a17b-20260216, Parasail, public 90-day aggregate· OpenRouter benchmark exports88.3%
524DeepSeek V4 ProOpenRouter harness, deepseek/deepseek-v4-pro-20260423, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports88.2%
525DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, Together, public 90-day aggregate· OpenRouter benchmark exports88.2%
526DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, PrimeIntellect, public 90-day aggregate· OpenRouter benchmark exports88.2%
527Kimi K2.7 CodeOpenRouter harness, moonshotai/kimi-k2.7-code-20260612, Phala, public 90-day aggregate· OpenRouter benchmark exports88.2%
528DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, Novita, public 90-day aggregate· OpenRouter benchmark exports88.2%
529Doubao Seed 2.0 Prohigh· Epoch AI Benchmarking Hub data88.2%
530GPT 5.2high· Epoch AI Benchmarking Hub data88.2%
531GPT-6 LunaOpenRouter harness, openai/gpt-6-luna-20260922, OpenAI, public 90-day aggregate· OpenRouter benchmark exports88.2%
532Doubao Seed 2.0 Litehigh· Epoch AI Benchmarking Hub data88.1%
533DeepSeek V4 Flash Vision ExpOpenRouter harness, deepseek/deepseek-v4-flash-vision-exp-20260821, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports88.1%
534DeepSeek V4 ProOpenRouter harness, deepseek/deepseek-v4-pro-20260423, Baidu, public 90-day aggregate· OpenRouter benchmark exports88.1%
535DeepSeek V4 FlashDeepSeek V4 official model card · Think Max· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.1%
536GLM 5.3DeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.1%
537GLM 5.3Pass@1 / max effort· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.1%
538Claude Fable 5.1OpenRouter harness, anthropic/claude-fable-5.1-20260831, Google, public 90-day aggregate· OpenRouter benchmark exports88.1%
539Mistral Large 4high· Epoch AI Benchmarking Hub data88.1%
540Claude Opus 5OpenRouter harness, anthropic/claude-opus-5-20260723, auto-routing, public 90-day aggregate· OpenRouter benchmark exports88.1%
541DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports88%
542DeepSeek V4 Flash Vision ExpOpenRouter harness, deepseek/deepseek-v4-flash-vision-exp-20260821, auto-routing, public 90-day aggregate· OpenRouter benchmark exports88%
543DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, Modal, public 90-day aggregate· OpenRouter benchmark exports88%
544GPT-5.4 minixhigh · published setting· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88%
545Kimi K2.7 CodeOpenRouter harness, moonshotai/kimi-k2.7-code-20260612, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports88%
546DeepSeek V4 ProOpenRouter harness, deepseek/deepseek-v4-pro-20260423, auto-routing, public 90-day aggregate· OpenRouter benchmark exports88%
547GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, OpenInference, public 90-day aggregate· OpenRouter benchmark exports87.9%
548GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, DigitalOcean, public 90-day aggregate· OpenRouter benchmark exports87.9%
549GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna-20260709, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports87.9%
550Qwen3.7 PlusOpenRouter harness, qwen/qwen3.7-plus-20260602, auto-routing, public 90-day aggregate· OpenRouter benchmark exports87.9%

Showing 501–550 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)