Calibration
AA-GPQA Diamond → HLE-Verified
HLE-Verified is estimated from AA-GPQA Diamond with a offset logistic curve fitted on 6 models measured on both: y = 0.1532 + (0.5458 − 0.1532) / (1 + exp(−200.00·(x − 0.9130))), R² = 1.00, cross-validated error 2.0 pp. It is used for 150 estimates.
| Estimated model | AA-GPQA Diamond | HLE-Verified | Source |
|---|---|---|---|
| A.X K2 | 85.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Apodex 1.1 | 86.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Apodex 1.1 Mini | 86.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Celeris-1 | 63.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Claude 3 Haiku | 37.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Claude 3 Opus | 48.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Claude 4.1 Opus Thinking | 80.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Claude 4 Sonnet | 68.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Claude Opus 4.5 | 81.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| Claude Opus 4.6 | 84.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| Claude Opus 4.7 | 88.5% | 15.5% | estimated ± 2.0 pp, low confidence |
| Command A+ | 76.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| DeepSeek-R1 | 81.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| DeepSeek R1 Distill Qwen 32B | 61.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| DeepSeek V3 | 55.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| DeepSeek V3 0324 | 65.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| DeepSeek V3.1 | 73.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| DeepSeek V3.1 (Reasoning) | 77.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| DeepSeek V3.2 | 75.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Exaone 4.0 1.2B | 42.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Exaone 4.0 32B | 62.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemini 1.0 Pro | 27.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemini 1.5 Pro | 58.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemini 2.5 Flash | 68.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemini 2.5 Pro | 84.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemini 3.5 Flash-Lite | 83.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemini 3 Flash | 81.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemma 3 27B | 42.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemma 4 12B | 75.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemma 4 26B A4B | 79.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemma 4 31B | 85.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemma 4 E2B | 43.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Gemma 4 E4B | 57.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| GLM-4.5-Air | 73.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| GLM-4.6 | 63.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| GLM-4.7 | 85.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| GLM-5 | 82.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| GLM-5.1 | 86.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| GLM-5.2 | 89.5% | 16.4% | estimated ± 2.0 pp, low confidence |
| GLM-5.3 | 91.7% | 42.3% | estimated ± 2.0 pp, medium confidence |
| GLM-5.3-Flash | 91.2% | 32.9% | estimated ± 2.0 pp, medium confidence |
| GLM-5-Turbo | 84.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| GLM-5V-Turbo | 80.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-4.1 | 66.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-4.1 mini | 66.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-4.1 nano | 51.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-4o | 54.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-4o mini | 42.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-5.1 | 87.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-5.1-Codex | 86.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-5.1-Codex-Max | 86.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-5.2-Codex | 89.9% | 17.6% | estimated ± 2.0 pp, low confidence |
| GPT-5.3 Codex | 91.5% | 38.7% | estimated ± 2.0 pp, medium confidence |
| GPT-5 (high) | 85.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-5 (medium) | 84.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-OSS 120B | 78.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| GPT-OSS 20B | 68.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Granite-4.0-350M | 26.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Granite-4.0-H-1B | 26.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Granite-4.0-H-350M | 25.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Granite 4.2 30B | 64.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Granite 4.2 3B | 55.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Granite 4.2 8B | 63.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Grok 4 | 87.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Grok 4.1 Fast | 63.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Grok 4.1 Fast (Reasoning) | 85.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Grok 4.3 | 90.1% | 18.6% | estimated ± 2.0 pp, low confidence |
| Grok 4 Fast (Reasoning) | 84.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Grok Code Fast 1 | 72.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Hy3 | 89.7% | 16.8% | estimated ± 2.0 pp, low confidence |
| Hy3 Preview | 89.7% | 16.8% | estimated ± 2.0 pp, low confidence |
| Inkling | 87.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| K-Exaone | 78.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| K-EXAONE 2.0 | 82.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Kimi K2.6 | 91.1% | 31.0% | estimated ± 2.0 pp, medium confidence |
| Kimi K2 | 76.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| Kimi K2.5 | 87.9% | 15.4% | estimated ± 2.0 pp, low confidence |
| Kimi K2.5 (Reasoning) | 87.9% | 15.4% | estimated ± 2.0 pp, low confidence |
| Kimi K2.7 Code | 89.6% | 16.6% | estimated ± 2.0 pp, low confidence |
| LFM2.5-2.6B | 55.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| LFM2.5-8B-A1B | 51.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| LFM2.5-VL-1.6B-Extract | 28.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Ling 2.6 Flash | 59.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Ling 3.0 Flash | 85.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Ling 3.0 Flash FP8 | 85.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Ling 3.0 Flash VL | 86.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| Ling 3.0 Tiny | 73.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Llama 3.1 405B | 51.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Llama 4 Maverick | 67.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Llama 4 Scout | 58.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| MiMo-V2.5-Pro | 86.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| MiMo-V2-Flash | 65.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| MiMo-V2-Omni | 82.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| MiMo-V2-Pro | 87.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| MiniCPM5-2B | 70.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| MiniMax M2.7 | 87.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| MiniMax M3 | 92.9% | 53.0% | estimated ± 2.0 pp, medium confidence |
| Mistral Large 2 | 48.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| Mistral Large 3 | 68.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| Mistral Medium 3 | 57.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Mistral Medium 3.5 128B | 74.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Mistral Small 4 | 76.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Mistral Small 4 (Reasoning) | 76.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Muse Glimmer 30B | 83.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Muse Spark 1.1 | 89.8% | 17.2% | estimated ± 2.0 pp, low confidence |
| Muse Spark 1.2 | 90.4% | 20.8% | estimated ± 2.0 pp, low confidence |
| Muse Spark 1.3 | 93.5% | 54.1% | estimated ± 2.0 pp, medium confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 74.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Nemotron 3 Nano 30B | 75.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Nemotron 3 Nano Omni 30B A3B | 46.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Nemotron 3 Super 100B | 80.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| Nemotron 3 Ultra | 86.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Nemotron Ultra 253B | 72.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| North Mini Code | 75.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Nova Pro | 49.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| o1 | 74.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| o1-preview | 76.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| o3 | 82.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| o3-mini | 74.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| o3-pro | 84.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Phi-4 | 57.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Phi-4 Multimodal Instruct | 31.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Quasar 438B | 73.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen2.5 Coder 32B Instruct | 41.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3.5-122B-A10B | 85.7% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3.5-27B | 85.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3.5-35B-A3B | 84.5% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3.5 397B | 86.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3.5 397B (Reasoning) | 86.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3.6-27B | 84.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3.6-35B-A3B | 84.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen 3.6 Max (preview) | 88.8% | 15.6% | estimated ± 2.0 pp, low confidence |
| Qwen3.6 Plus | 88.2% | 15.4% | estimated ± 2.0 pp, low confidence |
| Qwen3.7 Max | 92.3% | 49.9% | estimated ± 2.0 pp, medium confidence |
| Qwen3.7 Plus | 90.0% | 18.0% | estimated ± 2.0 pp, low confidence |
| Qwen3.8-27B | 90.5% | 21.9% | estimated ± 2.0 pp, low confidence |
| Qwen3.8-Flash-Next | 92.3% | 49.9% | estimated ± 2.0 pp, medium confidence |
| Qwen3.8 Max Preview | 92.8% | 52.7% | estimated ± 2.0 pp, medium confidence |
| Qwen3 Max | 76.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3-Omni-30B-A3B-Instruct | 62.0% | 15.3% | estimated ± 2.0 pp, low confidence |
| Qwen3-Omni-30B-A3B-Thinking | 72.6% | 15.3% | estimated ± 2.0 pp, low confidence |
| Sarvam 105B | 73.8% | 15.3% | estimated ± 2.0 pp, low confidence |
| Sarvam 30B | 63.3% | 15.3% | estimated ± 2.0 pp, low confidence |
| Solar Pro 2 | 56.1% | 15.3% | estimated ± 2.0 pp, low confidence |
| Solar Pro 3 | 72.4% | 15.3% | estimated ± 2.0 pp, low confidence |
| Solar Pro 4 | 89.1% | 15.8% | estimated ± 2.0 pp, low confidence |
| Step 3.7 Flash | 80.9% | 15.3% | estimated ± 2.0 pp, low confidence |
| Trinity-Large-Preview | 75.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| Trinity-Large-Thinking | 75.2% | 15.3% | estimated ± 2.0 pp, low confidence |
| Ultravox v0.6 Llama 3.3 70B | 49.8% | 15.3% | estimated ± 2.0 pp, low confidence |