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
651Inkling SmallOpenRouter harness, thinkingmachines/inkling-small-20260730, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports86.8%
652Qwen3.8 MaxOpenRouter harness, qwen/qwen3.8-2.4t-a95b-20260812, Modal, public 90-day aggregate· OpenRouter benchmark exports86.8%
653GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, InferenceNet, public 90-day aggregate· OpenRouter benchmark exports86.8%
654Kimi K2.6OpenRouter harness, moonshotai/kimi-k2.6-20260420, Chutes, public 90-day aggregate· OpenRouter benchmark exports86.8%
655Nemotron 3.5 Super VLOpenRouter harness, nvidia/nemotron-3.5-super-vl-20261015, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.8%
656GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Inceptron, public 90-day aggregate· OpenRouter benchmark exports86.8%
657MiniMax M3.1 FlashOpenRouter harness, minimax/minimax-m3.1-flash-20261019, SambaNova, public 90-day aggregate· OpenRouter benchmark exports86.8%
658Claude Haiku 5.5OpenRouter harness, anthropic/claude-haiku-5.5-20261007, Google, public 90-day aggregate· OpenRouter benchmark exports86.8%
659GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna-20260709, OpenAI, public 90-day aggregate· OpenRouter benchmark exports86.8%
660Qwen3.5 397B A17BOpenRouter harness, qwen/qwen3.5-397b-a17b-20260216, Novita, public 90-day aggregate· OpenRouter benchmark exports86.7%
661Claude Opus 4.6OpenRouter harness, anthropic/claude-4.6-opus-20260205, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.7%
662Claude Haiku 5.5OpenRouter harness, anthropic/claude-haiku-5.5-20261007, Azure, public 90-day aggregate· OpenRouter benchmark exports86.7%
663GPT 5.1OpenRouter harness, openai/gpt-5.1-20251113, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.7%
664GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Novita, public 90-day aggregate· OpenRouter benchmark exports86.7%
665Claude Opus 4.6OpenRouter harness, anthropic/claude-4.6-opus-20260205, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports86.7%
666Claude Opus 4.6OpenRouter harness, anthropic/claude-4.6-opus-20260205, Anthropic, public 90-day aggregate· OpenRouter benchmark exports86.7%
667Claude Haiku 5.5OpenRouter harness, anthropic/claude-haiku-5.5-20261007, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.6%
668Claude Opus 4.5OpenRouter harness, anthropic/claude-4.5-opus-20251124, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports86.6%
669Qwen3.5 397B A17BOpenRouter harness, qwen/qwen3.5-397b-a17b-20260216, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports86.6%
670GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, InferenceNet, public 90-day aggregate· OpenRouter benchmark exports86.6%
671Claude Opus 4.6OpenRouter harness, anthropic/claude-4.6-opus-20260205, Google, public 90-day aggregate· OpenRouter benchmark exports86.6%
672Qwen3.5 122B-A10B· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.86.6%
673MiniMax M3.1 FlashOpenRouter harness, minimax/minimax-m3.1-flash-20261019, Relace, public 90-day aggregate· OpenRouter benchmark exports86.5%
674DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, NextBit, public 90-day aggregate· OpenRouter benchmark exports86.5%
675Claude Opus 4.6OpenRouter harness, anthropic/claude-4.6-opus-20260205, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports86.5%
676DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, Cohere, public 90-day aggregate· OpenRouter benchmark exports86.5%
677Kimi K2.5OpenRouter harness, moonshotai/kimi-k2.5-0127, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.4%
678Qwen3.5 397B A17BOpenRouter harness, qwen/qwen3.5-397b-a17b-20260216, StreamLake, public 90-day aggregate· OpenRouter benchmark exports86.4%
679MiMo V2.5 ProOpenRouter harness, xiaomi/mimo-v2.5-pro-20260422, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports86.4%
680Claude Opus 4.5OpenRouter harness, anthropic/claude-4.5-opus-20251124, Anthropic, public 90-day aggregate· OpenRouter benchmark exports86.4%
681MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports86.4%
682GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Phala, public 90-day aggregate· OpenRouter benchmark exports86.4%
683Claude Opus 4.7max· Epoch AI Benchmarking Hub data86.4%
684Gemini 3.5 Flashminimal· Epoch AI Benchmarking Hub data86.4%
685Gemini 3.6 Flashlow· Epoch AI Benchmarking Hub data86.4%
686Qwen3.5 397B A17Bnone· Epoch AI Benchmarking Hub data86.4%
687Claude Opus 4.5OpenRouter harness, anthropic/claude-4.5-opus-20251124, Azure, public 90-day aggregate· OpenRouter benchmark exports86.4%
688MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, Novita, public 90-day aggregate· OpenRouter benchmark exports86.3%
689DeepSeek V4 Flash Vision ExpOpenRouter harness, deepseek/deepseek-v4-flash-vision-exp-20260821, Fireworks, public 90-day aggregate· OpenRouter benchmark exports86.3%
690Claude Opus 4.5OpenRouter harness, anthropic/claude-4.5-opus-20251124, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.3%
691Qwen3.8 MaxOpenRouter harness, qwen/qwen3.8-2.4t-a95b-20260812, Alibaba, public 90-day aggregate· OpenRouter benchmark exports86.3%
692Inkling SmallOpenRouter harness, thinkingmachines/inkling-small-20260730, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.3%
693DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, Relace, public 90-day aggregate· OpenRouter benchmark exports86.3%
694Claude Opus 4.5OpenRouter harness, anthropic/claude-4.5-opus-20251124, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports86.2%
695Qwen3.8 MaxOpenRouter harness, qwen/qwen3.8-2.4t-a95b-20260812, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports86.2%
696DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, Baidu, public 90-day aggregate· OpenRouter benchmark exports86.2%
697MiniMax M3.1 FlashOpenRouter harness, minimax/minimax-m3.1-flash-20261019, InferenceNet, public 90-day aggregate· OpenRouter benchmark exports86.2%
698MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, GMICloud, public 90-day aggregate· OpenRouter benchmark exports86.2%
699DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, OpenInference, public 90-day aggregate· OpenRouter benchmark exports86.2%
700GLM 5.1GLM-5.1 official release table· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86.2%

Showing 651–700 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)