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
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 | |
|---|---|---|
| 1301 | Claude Haiku 4.532k· Epoch AI Benchmarking Hub data | 71.2% |
| 1302 | DeepSeek Chat· Epoch AI Benchmarking Hub data | 71.2% |
| 1303 | GLM 5.2none· Epoch AI Benchmarking Hub data | 71.2% |
| 1304 | DeepSeek V4 FlashDeepSeek V4 official model card · Non-Think· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 71.2% |
| 1305 | Qwen3 VL 235BOpenRouter harness, qwen/qwen3-vl-235b-a22b-instruct, Alibaba, public 90-day aggregate· OpenRouter benchmark exports | 71.1% |
| 1306 | Nemotron 3.5 LightningOpenRouter harness, nvidia/nemotron-3.5-lightning-20260807, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports | 71.1% |
| 1307 | Qwen3 Next 80B A3B InstructOpenRouter harness, qwen/qwen3-next-80b-a3b-instruct-2509, Novita, public 90-day aggregate· OpenRouter benchmark exports | 71% |
| 1308 | Gemma 4 26B A4BOpenRouter harness, google/gemma-4-26b-a4b-it-20260403, AkashML, public 90-day aggregate· OpenRouter benchmark exports | 71% |
| 1309 | Qwen3 VL 235BOpenRouter harness, qwen/qwen3-vl-235b-a22b-instruct, PrimeIntellect, public 90-day aggregate· OpenRouter benchmark exports | 70.7% |
| 1310 | Qwen 3 235B A22B· Epoch AI Benchmarking Hub data | 70.7% |
| 1311 | Qwen3 VL 235BOpenRouter harness, qwen/qwen3-vl-235b-a22b-instruct, Novita, public 90-day aggregate· OpenRouter benchmark exports | 70.6% |
| 1312 | Nemotron 3.5 LightningOpenRouter harness, nvidia/nemotron-3.5-lightning-20260807, auto-routing, public 90-day aggregate· OpenRouter benchmark exports | 70.6% |
| 1313 | Gemma 4 26B A4BOpenRouter harness, google/gemma-4-26b-a4b-it-20260403, auto-routing, public 90-day aggregate· OpenRouter benchmark exports | 70.6% |
| 1314 | Qwen3 Next 80B A3B InstructOpenRouter harness, qwen/qwen3-next-80b-a3b-instruct-2509, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports | 70.5% |
| 1315 | GLM 4.6OpenRouter harness, z-ai/glm-4.6, Venice, public 90-day aggregate· OpenRouter benchmark exports | 70.5% |
| 1316 | GPT-5 nanoOpenRouter harness, openai/gpt-5-nano-2025-08-07, auto-routing, public 90-day aggregate· OpenRouter benchmark exports | 70.5% |
| 1317 | DeepSeek V4 ProOpenRouter harness, deepseek/deepseek-v4-pro-20260423, Venice, public 90-day aggregate· OpenRouter benchmark exports | 70.4% |
| 1318 | GLM 4.7OpenRouter harness, z-ai/glm-4.7-20251222, Venice, public 90-day aggregate· OpenRouter benchmark exports | 70.3% |
| 1319 | Qwen3.5 9BOpenRouter harness, qwen/qwen3.5-9b-20260310, Darkbloom, public 90-day aggregate· OpenRouter benchmark exports | 70.2% |
| 1320 | Qwen3 30B A3B2507, thinking· Epoch AI Benchmarking Hub data | 70.1% |
| 1321 | GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, Google, public 90-day aggregate· OpenRouter benchmark exports | 70% |
| 1322 | Claude Sonnet 4no extended thinking· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 70% |
| 1323 | Qwen3.6 35B A3BOpenRouter harness, qwen/qwen3.6-35b-a3b-20260415, Io Net, public 90-day aggregate· OpenRouter benchmark exports | 69.6% |
| 1324 | Qwen3 VL 235BOpenRouter harness, qwen/qwen3-vl-235b-a22b-instruct, auto-routing, public 90-day aggregate· OpenRouter benchmark exports | 69.4% |
| 1325 | GPT-5 nanohigh· Epoch AI Benchmarking Hub data | 69.4% |
| 1326 | Claude Opus 4· Epoch AI Benchmarking Hub data | 69.2% |
| 1327 | Qwen3 VL 235BOpenRouter harness, qwen/qwen3-vl-235b-a22b-instruct, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports | 69.1% |
| 1328 | Gemma 4 31B ITOpenRouter harness, google/gemma-4-31b-it-20260402, Friendli, public 90-day aggregate· OpenRouter benchmark exports | 69.1% |
| 1329 | GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports | 69.1% |
| 1330 | GPT-5 nanoOpenRouter harness, openai/gpt-5-nano-2025-08-07, Azure, public 90-day aggregate· OpenRouter benchmark exports | 69% |
| 1331 | GPT-4.5 Preview· Epoch AI Benchmarking Hub data | 68.7% |
| 1332 | GPT OSS 20bOpenRouter harness, openai/gpt-oss-20b, DekaLLM, public 90-day aggregate· OpenRouter benchmark exports | 68.7% |
| 1333 | GPT-5 nanoOpenRouter harness, openai/gpt-5-nano-2025-08-07, OpenAI, public 90-day aggregate· OpenRouter benchmark exports | 68.4% |
| 1334 | DeepSeek V3.1OpenRouter harness, deepseek/deepseek-chat-v3.1, SambaNova, public 90-day aggregate· OpenRouter benchmark exports | 68.3% |
| 1335 | o3 Minilow· Epoch AI Benchmarking Hub data | 68.2% |
| 1336 | GPT OSS 120BOpenRouter harness, openai/gpt-oss-120b, Mancer 2, public 90-day aggregate· OpenRouter benchmark exports | 68.2% |
| 1337 | Gemma 4 31B ITOpenRouter harness, google/gemma-4-31b-it-20260402, OpenInference, public 90-day aggregate· OpenRouter benchmark exports | 68.2% |
| 1338 | Nemotron 3 Nano 30B A3BOpenRouter harness, nvidia/nemotron-3-nano-30b-a3b, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports | 68.1% |
| 1339 | GPT OSS 120Bofficial 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. | 68.1% |
| 1340 | MiniMax M2.5OpenRouter harness, minimax/minimax-m2.5-20260211, Venice, public 90-day aggregate· OpenRouter benchmark exports | 68.1% |
| 1341 | GPT OSS 20bOpenRouter harness, openai/gpt-oss-20b, AkashML, public 90-day aggregate· OpenRouter benchmark exports | 67.8% |
| 1342 | DeepSeek V3OpenRouter harness, deepseek/deepseek-chat-v3-0324, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports | 67.7% |
| 1343 | DeepSeek V3· Epoch AI Benchmarking Hub data | 67.6% |
| 1344 | Grok 3beta· Epoch AI Benchmarking Hub data | 67.6% |
| 1345 | GPT OSS 20bOpenRouter harness, openai/gpt-oss-20b, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports | 67.4% |
| 1346 | GPT OSS 20bOpenRouter harness, openai/gpt-oss-20b, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports | 67.4% |
| 1347 | GPT-5 nanomedium· Epoch AI Benchmarking Hub data | 67.4% |
| 1348 | GPT OSS 20bOpenRouter harness, openai/gpt-oss-20b, Parasail, public 90-day aggregate· OpenRouter benchmark exports | 67.1% |
| 1349 | GPT OSS 120B· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it. | 67.1% |
| 1350 | GPT OSS 120Bofficial 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. | 67.1% |
Showing 1,301–1,350 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)