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
1451Qwen3 14BOpenRouter harness, qwen/qwen3-14b-04-28, auto-routing, public 90-day aggregate· OpenRouter benchmark exports59.2%
1452GLM 4.7 FlashOpenRouter harness, z-ai/glm-4.7-flash-20260119, Novita, public 90-day aggregate· OpenRouter benchmark exports59.1%
1453GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, Cerebras, public 90-day aggregate· OpenRouter benchmark exports58.8%
1454GPT OSS 20bOpenRouter harness, openai/gpt-oss-20b, Novita, public 90-day aggregate· OpenRouter benchmark exports58.5%
1455GPT OSS 20bofficial model card Table 3 · reasoning low · with tools; Table 3 does not name the exact tool · pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.58%
1456GPT OSS 20bOpenRouter harness, openai/gpt-oss-20b, Google, public 90-day aggregate· OpenRouter benchmark exports57.7%
1457GPT-5 nanolow· Epoch AI Benchmarking Hub data57.6%
1458gemini-1.5-pro-002· Epoch AI Benchmarking Hub data57.2%
1459Llama 4 Scoutinstruction-tuned table· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.57.2%
1460Gemini 2.0 Flash Experimentalthinking· Epoch AI Benchmarking Hub data57.1%
1461GPT OSS 20b· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.56.8%
1462GPT OSS 20bofficial model card Table 3 · reasoning low · no tools · pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56.8%
1463Qwen3 8B· Epoch AI Benchmarking Hub data56.8%
1464Qwen Max· Epoch AI Benchmarking Hub data56.1%
1465Gemini 2.5 Flash-LiteOpenRouter harness, google/gemini-2.5-flash-lite, Google, public 90-day aggregate· OpenRouter benchmark exports56.1%
1466Magistral Small 25092506· Epoch AI Benchmarking Hub data56.1%
1467Phi 4· Epoch AI Benchmarking Hub data56.1%
1468DeepSeek R1 Distill Llama 70b· Epoch AI Benchmarking Hub data55.7%
1469Qwen3 32BOpenRouter harness, qwen/qwen3-32b-04-28, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports55.7%
1470Qwen3 30B A3B Instruct· Epoch AI Benchmarking Hub data55.6%
1471GPT 5.4 Nanonone· Epoch AI Benchmarking Hub data55.6%
1472Claude 3.5 Sonnet· Epoch AI Benchmarking Hub data55.3%
1473Gemini 2.5 Flash-LiteOpenRouter harness, google/gemini-2.5-flash-lite, auto-routing, public 90-day aggregate· OpenRouter benchmark exports55.2%
1474GLM 5OpenRouter harness, z-ai/glm-5-20260211, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports54.9%
1475Qwen3 32BOpenRouter harness, qwen/qwen3-32b-04-28, Tenstorrent, public 90-day aggregate· OpenRouter benchmark exports54.7%
1476DeepSeek V4 ProOpenRouter harness, deepseek/deepseek-v4-pro-20260423, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports54.3%
1477Qwen3 32Bnone· Epoch AI Benchmarking Hub data54.1%
1478Qwen3 VL 8B InstructOpenRouter harness, qwen/qwen3-vl-8b-instruct, PrimeIntellect, public 90-day aggregate· OpenRouter benchmark exports53.9%
1479Grok 2· Epoch AI Benchmarking Hub data53.8%
1480DeepSeek ChatOpenRouter harness, deepseek/deepseek-chat-v3, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports53.7%
1481Qwen3 14Bnone· Epoch AI Benchmarking Hub data53.4%
1482GPT OSS 20blow· Epoch AI Benchmarking Hub data53.2%
1483GPT 4oOpenRouter harness, openai/gpt-4o-2024-08-06, OpenAI, public 90-day aggregate· OpenRouter benchmark exports52.8%
1484GPT 4oOpenRouter harness, openai/gpt-4o, OpenAI, public 90-day aggregate· OpenRouter benchmark exports52.7%
1485Qwen3 4B· Epoch AI Benchmarking Hub data52.3%
1486DeepSeek V3OpenRouter harness, deepseek/deepseek-chat-v3-0324, GMICloud, public 90-day aggregate· OpenRouter benchmark exports52.1%
1487GPT 4oOpenRouter harness, openai/gpt-4o-2024-08-06, auto-routing, public 90-day aggregate· OpenRouter benchmark exports52%
1488Llama 4 Scout· Epoch AI Benchmarking Hub data51.8%
1489GPT 4oOpenRouter harness, openai/gpt-4o, Azure, public 90-day aggregate· OpenRouter benchmark exports51.7%
1490GPT 4oOpenRouter harness, openai/gpt-4o-2024-08-06, Azure, public 90-day aggregate· OpenRouter benchmark exports51.5%
1491Mistral Large2411· Epoch AI Benchmarking Hub data51.3%
1492DeepSeek V3OpenRouter harness, deepseek/deepseek-chat-v3-0324, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports51.3%
1493GLM 4.7 FlashOpenRouter harness, z-ai/glm-4.7-flash-20260119, auto-routing, public 90-day aggregate· OpenRouter benchmark exports51.1%
1494GPT 4oOpenRouter harness, openai/gpt-4o, auto-routing, public 90-day aggregate· OpenRouter benchmark exports51.1%
1495Llama 3.1 405B Instruct· Epoch AI Benchmarking Hub data50.9%
1496GPT-4.1 nanoOpenRouter harness, openai/gpt-4.1-nano-2025-04-14, OpenAI, public 90-day aggregate· OpenRouter benchmark exports50.5%
1497Ling 3.0 Flash FinOpenRouter harness, inclusionai/ling-3.0-flash-fin-20260827, auto-routing, public 90-day aggregate· OpenRouter benchmark exports50.5%
1498Qwen3 30B A3Bnone· Epoch AI Benchmarking Hub data50.4%
1499O1 Preview· Epoch AI Benchmarking Hub data50.3%
1500GPT-4.1 nanoOpenRouter harness, openai/gpt-4.1-nano-2025-04-14, Azure, public 90-day aggregate· OpenRouter benchmark exports50%

Showing 1,451–1,500 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)