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
351GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, DigitalOcean, public 90-day aggregate· OpenRouter benchmark exports89.9%
352DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, Wafer, public 90-day aggregate· OpenRouter benchmark exports89.9%
353Claude Fable 5.1OpenRouter harness, anthropic/claude-fable-5.1-20260831, auto-routing, public 90-day aggregate· OpenRouter benchmark exports89.9%
354GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, Wafer, public 90-day aggregate· OpenRouter benchmark exports89.9%
355Kimi K2.6OpenRouter harness, moonshotai/kimi-k2.6-20260420, Novita, public 90-day aggregate· OpenRouter benchmark exports89.9%
356Claude Opus 5.5OpenRouter harness, anthropic/claude-opus-5.5-20260921, Google, public 90-day aggregate· OpenRouter benchmark exports89.9%
357Claude Opus 5.5OpenRouter harness, anthropic/claude-opus-5.5-20260921, Azure, public 90-day aggregate· OpenRouter benchmark exports89.9%
358Claude Sonnet 4.6adaptive thinking · max effort · 10 trials· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.9%
359DeepSeek V4 Flash 0731DeepSeek 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.89.9%
360GLM 5.1· Epoch AI Benchmarking Hub data89.9%
361GPT 5.4high· Epoch AI Benchmarking Hub data89.9%
362GPT-5.6 Sollow· Epoch AI Benchmarking Hub data89.9%
363DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, DigitalOcean, public 90-day aggregate· OpenRouter benchmark exports89.9%
364Claude Opus 4.7OpenRouter harness, anthropic/claude-4.7-opus-20260416, Google, public 90-day aggregate· OpenRouter benchmark exports89.9%
365DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, GMICloud, public 90-day aggregate· OpenRouter benchmark exports89.9%
366DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, Relace, public 90-day aggregate· OpenRouter benchmark exports89.8%
367DeepSeek V4.1 Flashmax· Epoch AI Benchmarking Hub data89.8%
368DeepSeek V4 Flash Vision ExpOpenRouter harness, deepseek/deepseek-v4-flash-vision-exp-20260821, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports89.8%
369Muse Spark· Epoch AI Benchmarking Hub data89.8%
370DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, Ionstream, public 90-day aggregate· OpenRouter benchmark exports89.8%
371Kimi K2.6OpenRouter harness, moonshotai/kimi-k2.6-20260420, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports89.8%
372DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports89.8%
373Claude Opus 4.7OpenRouter harness, anthropic/claude-4.7-opus-20260416, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports89.8%
374MiMo V2.6 ProOpenRouter harness, xiaomi/mimo-v2.6-pro-20260921, StreamLake, public 90-day aggregate· OpenRouter benchmark exports89.8%
375DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, AkashML, public 90-day aggregate· OpenRouter benchmark exports89.8%
376Claude Opus 4.8OpenRouter harness, anthropic/claude-4.8-opus-20260528, Anthropic, public 90-day aggregate· OpenRouter benchmark exports89.8%
377Kimi K2.6OpenRouter harness, moonshotai/kimi-k2.6-20260420, Moonshot AI, public 90-day aggregate· OpenRouter benchmark exports89.8%
378Kimi K2.6OpenRouter harness, moonshotai/kimi-k2.6-20260420, Decart, public 90-day aggregate· OpenRouter benchmark exports89.8%
379GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna-pro-20260709, OpenAI, public 90-day aggregate, pro· OpenRouter benchmark exports89.7%
380Claude Opus 4.7OpenRouter harness, anthropic/claude-4.7-opus-20260416, auto-routing, public 90-day aggregate· OpenRouter benchmark exports89.7%
381DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, Crusoe, public 90-day aggregate· OpenRouter benchmark exports89.7%
382Muse Spark OpenOpenRouter harness, meta/muse-spark-open-20261015, Together, public 90-day aggregate· OpenRouter benchmark exports89.7%
383Claude Opus 5OpenRouter harness, anthropic/claude-opus-5-20260723, Anthropic, public 90-day aggregate· OpenRouter benchmark exports89.7%
384Muse Spark OpenOpenRouter harness, meta/muse-spark-open-20261015, Nebius, public 90-day aggregate· OpenRouter benchmark exports89.7%
385MiniMax M2.7OpenRouter harness, minimax/minimax-m2.7-20260318, Groq, public 90-day aggregate· OpenRouter benchmark exports89.7%
386Hy4 previewOpenRouter harness, tencent/hy4-preview-20260827, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports89.7%
387Hy4 previewOpenRouter harness, tencent/hy4-preview-20260827, auto-routing, public 90-day aggregate· OpenRouter benchmark exports89.7%
388Claude Haiku 5.5max· Epoch AI Benchmarking Hub data89.6%
389DeepSeek V4 Promax· Epoch AI Benchmarking Hub data89.6%
390Kimi K2.7 CodeOpenRouter harness, moonshotai/kimi-k2.7-code-20260612, ModelRun, public 90-day aggregate· OpenRouter benchmark exports89.6%
391DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports89.6%
392Claude Opus 5.5OpenRouter harness, anthropic/claude-opus-5.5-20260921, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports89.6%
393Claude Opus 4.6SenseNova 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.89.6%
394Doubao Seed 2.1 Turbohigh· Epoch AI Benchmarking Hub data89.6%
395GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna-pro-20260709, auto-routing, public 90-day aggregate, pro· OpenRouter benchmark exports89.6%
396Hy4 previewOpenRouter harness, tencent/hy4-preview-20260827, GMICloud, public 90-day aggregate· OpenRouter benchmark exports89.6%
397DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-20260731, Parasail, public 90-day aggregate· OpenRouter benchmark exports89.6%
398DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports89.5%
399DeepSeek V4 FlashOpenRouter harness, deepseek/deepseek-v4-flash-20260423, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports89.5%
400DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, Phala, public 90-day aggregate· OpenRouter benchmark exports89.5%

Showing 351–400 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)