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
1251Claude Sonnet 4.5· Epoch AI Benchmarking Hub data73.7%
1252Gemini 3.1 Flash Liteminimal· Epoch AI Benchmarking Hub data73.7%
1253Claude Sonnet 4.6OpenRouter harness, anthropic/claude-4.6-sonnet-20260217, Google, public 90-day aggregate· OpenRouter benchmark exports73.7%
1254Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, Parasail, public 90-day aggregate· OpenRouter benchmark exports73.7%
1255Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, Google, public 90-day aggregate· OpenRouter benchmark exports73.6%
1256Kimi K2.6OpenRouter harness, moonshotai/kimi-k2.6-20260420, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports73.5%
1257GPT OSS 120Bofficial model card Table 3 · reasoning medium · 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.73.5%
1258Claude Sonnet 4.6OpenRouter harness, anthropic/claude-4.6-sonnet-20260217, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports73.5%
1259Claude Sonnet 4.6OpenRouter harness, anthropic/claude-4.6-sonnet-20260217, Anthropic, public 90-day aggregate· OpenRouter benchmark exports73.5%
1260Qwen3.6 35B A3BOpenRouter harness, qwen/qwen3.6-35b-a3b-20260415, Darkbloom, public 90-day aggregate· OpenRouter benchmark exports73.4%
1261Nemotron 3.5 LightningOpenRouter harness, nvidia/nemotron-3.5-lightning-20260807, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports73.3%
1262Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, Venice, public 90-day aggregate· OpenRouter benchmark exports73.3%
1263Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, Novita, public 90-day aggregate· OpenRouter benchmark exports73.3%
1264MiMo V2.6 FlashOpenRouter harness, xiaomi/mimo-v2.6-flash-20260921, Relace, public 90-day aggregate· OpenRouter benchmark exports73.2%
1265Claude Opus 4.1· Epoch AI Benchmarking Hub data73.2%
1266DeepSeek V4 Pronone· Epoch AI Benchmarking Hub data73.2%
1267Gemma 4 26B A4Bminimal· Epoch AI Benchmarking Hub data73.2%
1268GPT 5.2none· Epoch AI Benchmarking Hub data73.2%
1269MiMo V2.6 FlashOpenRouter harness, xiaomi/mimo-v2.6-flash-20260921, auto-routing, public 90-day aggregate· OpenRouter benchmark exports73.2%
1270Claude Haiku 4.5OpenRouter harness, anthropic/claude-4.5-haiku-20251001, auto-routing, public 90-day aggregate· OpenRouter benchmark exports73.2%
1271GPT OSS 120Bofficial model card Table 3 · reasoning medium · no tools · pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.1%
1272Claude Haiku 4.5launch-page setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73%
1273Claude Haiku 4.5OpenRouter harness, anthropic/claude-4.5-haiku-20251001, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports72.9%
1274DeepSeek V3.1OpenRouter harness, deepseek/deepseek-chat-v3.1, Mara, public 90-day aggregate· OpenRouter benchmark exports72.9%
1275Kimi K2.7 CodeOpenRouter harness, moonshotai/kimi-k2.7-code-20260612, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports72.8%
1276Qwen3 Next 80B A3B InstructOpenRouter harness, qwen/qwen3-next-80b-a3b-instruct-2509, Parasail, public 90-day aggregate· OpenRouter benchmark exports72.8%
1277Gemini 2.5 FlashOpenRouter harness, google/gemini-2.5-flash, auto-routing, public 90-day aggregate· OpenRouter benchmark exports72.8%
1278Claude Haiku 4.5OpenRouter harness, anthropic/claude-4.5-haiku-20251001, Azure, public 90-day aggregate· OpenRouter benchmark exports72.8%
1279Nemotron 3.5 LightningOpenRouter harness, nvidia/nemotron-3.5-lightning-20260807, Phala, public 90-day aggregate· OpenRouter benchmark exports72.7%
1280Claude Haiku 4.5OpenRouter harness, anthropic/claude-4.5-haiku-20251001, Anthropic, public 90-day aggregate· OpenRouter benchmark exports72.7%
1281MiMo V2.6 FlashOpenRouter harness, xiaomi/mimo-v2.6-flash-20260921, Cohere, public 90-day aggregate· OpenRouter benchmark exports72.7%
1282Kimi K2.7 CodeOpenRouter harness, moonshotai/kimi-k2.7-code-20260612, Tenstorrent, public 90-day aggregate· OpenRouter benchmark exports72.6%
1283Qwen3 Max· Epoch AI Benchmarking Hub data72.6%
1284Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, auto-routing, public 90-day aggregate· OpenRouter benchmark exports72.6%
1285DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, Krea, public 90-day aggregate· OpenRouter benchmark exports72.3%
1286GLM 5OpenRouter harness, z-ai/glm-5-20260211, Novita, public 90-day aggregate· OpenRouter benchmark exports72.3%
1287GLM 5OpenRouter harness, z-ai/glm-5-20260211, Z.AI, public 90-day aggregate· OpenRouter benchmark exports72.3%
1288GPT 5.4 Nanolow· Epoch AI Benchmarking Hub data72.2%
1289Claude Haiku 4.5OpenRouter harness, anthropic/claude-4.5-haiku-20251001, Google, public 90-day aggregate· OpenRouter benchmark exports72.1%
1290Qwen3 235B A22B Instruct 2507OpenRouter harness, qwen/qwen3-235b-a22b-07-25, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports72.1%
1291Qwen3 Next 80B A3B InstructOpenRouter harness, qwen/qwen3-next-80b-a3b-instruct-2509, Google, public 90-day aggregate· OpenRouter benchmark exports72.1%
1292Qwen3 Next 80B A3B InstructOpenRouter harness, qwen/qwen3-next-80b-a3b-instruct-2509, auto-routing, public 90-day aggregate· OpenRouter benchmark exports72%
1293MiMo V2.6 FlashOpenRouter harness, xiaomi/mimo-v2.6-flash-20260921, Makora, public 90-day aggregate· OpenRouter benchmark exports71.8%
1294GPT 5minimal· Epoch AI Benchmarking Hub data71.7%
1295GPT-5 miniminimal· Epoch AI Benchmarking Hub data71.7%
1296Qwen3 Next 80B A3B InstructOpenRouter harness, qwen/qwen3-next-80b-a3b-instruct-2509, Alibaba, public 90-day aggregate· OpenRouter benchmark exports71.7%
1297GPT-5 minimedium· Epoch AI Benchmarking Hub data71.7%
1298Seed OSS 36B Instruct· Epoch AI Benchmarking Hub data71.5%
1299GPT OSS 20bofficial model card Table 3 · reasoning high · no tools · pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.71.5%
1300Qwen3 VL 235BOpenRouter harness, qwen/qwen3-vl-235b-a22b-instruct, Parasail, public 90-day aggregate· OpenRouter benchmark exports71.4%

Showing 1,251–1,300 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)