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
701DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, Mancer 2, public 90-day aggregate· OpenRouter benchmark exports86.2%
702GPT 5high· Epoch AI Benchmarking Hub data86.2%
703GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Fireworks, public 90-day aggregate· OpenRouter benchmark exports86.2%
704DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, auto-routing, public 90-day aggregate· OpenRouter benchmark exports86.1%
705DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports86.1%
706Gemma 4 31B ITOpenRouter harness, google/gemma-4-31b-it-20260402, DekaLLM, public 90-day aggregate· OpenRouter benchmark exports86.1%
707GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, Venice, public 90-day aggregate· OpenRouter benchmark exports86.1%
708Claude Opus 4.532k· Epoch AI Benchmarking Hub data86%
709GLM 5GLM-5 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%
710Qwen3.6 FlashQwen3.6-35B-A3B official model card · hosted API name Qwen3.6-Flash· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86%
711Hy4 previewOpenRouter harness, tencent/hy4-preview-20260827, Io Net, public 90-day aggregate· OpenRouter benchmark exports86%
712GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, DigitalOcean, public 90-day aggregate· OpenRouter benchmark exports85.9%
713MiMo V2.5 ProOpenRouter harness, xiaomi/mimo-v2.5-pro-20260422, Xiaomi, public 90-day aggregate· OpenRouter benchmark exports85.9%
714GLM 5.3 FlashXOpenRouter harness, z-ai/glm-5.3-flashx-20260918, Novita, public 90-day aggregate· OpenRouter benchmark exports85.9%
715MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, PrimeIntellect, public 90-day aggregate· OpenRouter benchmark exports85.9%
716Claude Fable 5max· Epoch AI Benchmarking Hub data85.9%
717Gemini 3.6 Flashminimal· Epoch AI Benchmarking Hub data85.9%
718Qwen3.6 27B· Epoch AI Benchmarking Hub data85.9%
719MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, auto-routing, public 90-day aggregate· OpenRouter benchmark exports85.8%
720GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, PrimeIntellect, public 90-day aggregate· OpenRouter benchmark exports85.8%
721Claude Opus 4.5OpenRouter harness, anthropic/claude-4.5-opus-20251124, Google, public 90-day aggregate· OpenRouter benchmark exports85.8%
722Claude Sonnet 5OpenRouter harness, anthropic/claude-sonnet-5-20260630, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports85.7%
723MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, Minimax, public 90-day aggregate· OpenRouter benchmark exports85.7%
724GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, BaseTen, public 90-day aggregate· OpenRouter benchmark exports85.7%
725GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Baidu, public 90-day aggregate· OpenRouter benchmark exports85.7%
726DeepSeek V3.2 SpecialeSpeciale high-compute reasoning · temperature 1.0 · 128K context · pass@1· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.7%
727GLM 4.7· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.
Values from 2 sources
85.7%
728Step 3.5 FlashStep 3.5 technical report · Tables 12–13 · Python tool + PaCoRe· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.7%
729MiniMax M3.1 FlashOpenRouter harness, minimax/minimax-m3.1-flash-20261019, Wafer, public 90-day aggregate· OpenRouter benchmark exports85.7%
730GPT 5OpenRouter harness, openai/gpt-5-2025-08-07, auto-routing, public 90-day aggregate· OpenRouter benchmark exports85.7%
731GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Near AI, public 90-day aggregate· OpenRouter benchmark exports85.7%
732GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Z.AI, public 90-day aggregate· OpenRouter benchmark exports85.7%
733GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Decart, public 90-day aggregate· OpenRouter benchmark exports85.7%
734GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, GMICloud, public 90-day aggregate· OpenRouter benchmark exports85.6%
735GLM 5.1OpenRouter harness, z-ai/glm-5.1-20260406, Alibaba, public 90-day aggregate· OpenRouter benchmark exports85.6%
736GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Ambient, public 90-day aggregate· OpenRouter benchmark exports85.5%
737DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, Fireworks, public 90-day aggregate· OpenRouter benchmark exports85.5%
738Qwen3.5 27B· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.85.5%
739Qwen3.6 27BOpenRouter harness, qwen/qwen3.6-27b-20260422, Venice, public 90-day aggregate· OpenRouter benchmark exports85.5%
740Gemma 4 31B ITOpenRouter harness, google/gemma-4-31b-it-20260402, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports85.5%
741Claude Opus 4.516k· Epoch AI Benchmarking Hub data85.5%
742Qwen3.5 122B-A10BOpenRouter harness, qwen/qwen3.5-122b-a10b-20260224, Alibaba, public 90-day aggregate· OpenRouter benchmark exports85.5%
743MiMo V2.5 ProOpenRouter harness, xiaomi/mimo-v2.5-pro-20260422, Novita, public 90-day aggregate· OpenRouter benchmark exports85.5%
744Claude Sonnet 5OpenRouter harness, anthropic/claude-sonnet-5-20260630, auto-routing, public 90-day aggregate· OpenRouter benchmark exports85.5%
745Gemma 4 31B ITOpenRouter harness, google/gemma-4-31b-it-20260402, Crusoe, public 90-day aggregate· OpenRouter benchmark exports85.5%
746GLM 4.6OpenRouter harness, z-ai/glm-4.6, Z.AI, public 90-day aggregate· OpenRouter benchmark exports85.4%
747GLM 5.1OpenRouter harness, z-ai/glm-5.1-20260406, Nebius, public 90-day aggregate· OpenRouter benchmark exports85.4%
748GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports85.4%
749GLM 5.3OpenRouter harness, z-ai/glm-5.3-20260816, auto-routing, public 90-day aggregate· OpenRouter benchmark exports85.4%
750Claude Opus 4.8none· Epoch AI Benchmarking Hub data85.4%

Showing 701–750 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)