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 | |
|---|---|---|
| 201 | Kimi K3.1 ProOpenRouter harness, moonshotai/kimi-k3.1-pro-20261023, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports | 91.6% |
| 202 | MiniMax M3OpenRouter harness, minimax/minimax-m3-20260531, Together, public 90-day aggregate· OpenRouter benchmark exports | 91.6% |
| 203 | MiniMax M3OpenRouter harness, minimax/minimax-m3-20260531, StreamLake, public 90-day aggregate· OpenRouter benchmark exports | 91.5% |
| 204 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports | 91.5% |
| 205 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, Alibaba, public 90-day aggregate· OpenRouter benchmark exports | 91.5% |
| 206 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, Cloudflare, public 90-day aggregate· OpenRouter benchmark exports | 91.4% |
| 207 | Inkling 2 SmallOpenRouter harness, thinkingmachines/inkling-2-small-20261107, Nebius, public 90-day aggregate· OpenRouter benchmark exports | 91.4% |
| 208 | Inkling 2 SmallOpenRouter harness, thinkingmachines/inkling-2-small-20261107, Together, public 90-day aggregate· OpenRouter benchmark exports | 91.4% |
| 209 | GPT 5.2xhigh· Epoch AI Benchmarking Hub data | 91.4% |
| 210 | MiniMax M3OpenRouter harness, minimax/minimax-m3-20260531, Novita, public 90-day aggregate· OpenRouter benchmark exports | 91.3% |
| 211 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, Baidu, public 90-day aggregate· OpenRouter benchmark exports | 91.3% |
| 212 | Claude Opus 4.6Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 91.3% |
| 213 | Claude Opus 4.6Qwen3.8 Omni Flash 官方发布表 · max effort· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 91.3% |
| 214 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, Wafer, public 90-day aggregate· OpenRouter benchmark exports | 91.3% |
| 215 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, Phala, public 90-day aggregate· OpenRouter benchmark exports | 91.3% |
| 216 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, Alibaba, public 90-day aggregate· OpenRouter benchmark exports | 91.2% |
| 217 | Kimi K3OpenRouter harness, moonshotai/kimi-k3-20260715, Together, public 90-day aggregate· OpenRouter benchmark exports | 91.2% |
| 218 | Inkling 2 SmallOpenRouter harness, thinkingmachines/inkling-2-small-20261107, Cloudflare, public 90-day aggregate· OpenRouter benchmark exports | 91.2% |
| 219 | MiniMax M3OpenRouter harness, minimax/minimax-m3-20260531, GMICloud, public 90-day aggregate· OpenRouter benchmark exports | 91.2% |
| 220 | GLM 5.2setting not named by the source· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 91.2% |
| 221 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, Venice, public 90-day aggregate· OpenRouter benchmark exports | 91.2% |
| 222 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, GMICloud, public 90-day aggregate· OpenRouter benchmark exports | 91.1% |
| 223 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, Pearl Research, public 90-day aggregate· OpenRouter benchmark exports | 91.1% |
| 224 | GPT-6.1 TerraOpenRouter harness, openai/gpt-6.1-terra-20261026, OpenAI, public 90-day aggregate· OpenRouter benchmark exports | 91.1% |
| 225 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, DekaLLM, public 90-day aggregate· OpenRouter benchmark exports | 91.1% |
| 226 | Claude Opus 4.8max· Epoch AI Benchmarking Hub data | 91% |
| 227 | DeepSeek V4 Flash 0731max· Epoch AI Benchmarking Hub data | 91% |
| 228 | Kimi K3OpenRouter harness, moonshotai/kimi-k3-20260715, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports | 91% |
| 229 | Qwen3.8 Omni FlashQwen3.8 Omni Flash 官方发布表· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 91% |
| 230 | GPT-6.1 TerraOpenRouter harness, openai/gpt-6.1-terra-20261026, Azure, public 90-day aggregate· OpenRouter benchmark exports | 91% |
| 231 | Muse Spark OpenOpenRouter harness, meta/muse-spark-open-20261015, Modal, public 90-day aggregate· OpenRouter benchmark exports | 91% |
| 232 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, Novita, public 90-day aggregate· OpenRouter benchmark exports | 91% |
| 233 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, River, public 90-day aggregate· OpenRouter benchmark exports | 91% |
| 234 | GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna-pro-20260709, Azure, public 90-day aggregate, pro· OpenRouter benchmark exports | 91% |
| 235 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, auto-routing, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 236 | DeepSeek V4.1 FlashOpenRouter harness, deepseek/deepseek-v4.1-flash-20260910, NextBit, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 237 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, StreamLake, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 238 | DeepSeek V4 Pro 0813OpenRouter harness, deepseek/deepseek-v4-pro-20260813, Reka, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 239 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, Ambient, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 240 | Inkling 2 SmallOpenRouter harness, thinkingmachines/inkling-2-small-20261107, Modal, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 241 | Qwen3.7 MaxOpenRouter harness, qwen/qwen3.7-max-20260520, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 242 | DeepSeek V4 Prohigh· Epoch AI Benchmarking Hub data | 90.9% |
| 243 | GLM 5.3max· Epoch AI Benchmarking Hub data | 90.9% |
| 244 | MiniMax M3· Epoch AI Benchmarking Hub data | 90.9% |
| 245 | Qwen3.7 Max· Epoch AI Benchmarking Hub data | 90.9% |
| 246 | DeepSeek V4 Flash Vision ExpOpenRouter harness, deepseek/deepseek-v4-flash-vision-exp-20260821, GMICloud, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 247 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, StepFun, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
| 248 | DeepSeek V4.1 FlashDeepSeek 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. | 90.9% |
| 249 | DeepSeek V4.1 FlashPass@1 / max effort· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.9% |
| 250 | DeepSeek V4.1 Flash 1017OpenRouter harness, deepseek/deepseek-v4.1-flash-20261017, CoreWeave, public 90-day aggregate· OpenRouter benchmark exports | 90.9% |
Showing 201–250 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)