Calibration
CritPt → HLE
HLE is estimated from CritPt with a Hill curve fitted on 51 models measured on both: y = 0.1895 + (1.2000 − 0.1895)·x^0.55 / (0.75804^0.55 + x^0.55), R² = 0.55, cross-validated error 9.5 pp. It is used for 90 estimates.
| Estimated model | CritPt | HLE | Source |
|---|---|---|---|
| Apodex 1.1 Mini | 4.6% | 36.7% | estimated ± 9.5 pp, medium confidence |
| Celeris-1 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Claude 3 Haiku | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Claude 4.1 Opus Thinking | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Claude 4 Sonnet | 1.1% | 27.8% | estimated ± 9.5 pp, medium confidence |
| Claude Opus 4.7 | 5.1% | 37.5% | estimated ± 9.5 pp, medium confidence |
| Command A+ | 0.3% | 23.5% | estimated ± 9.5 pp, medium confidence |
| DeepSeek-R1 | 1.4% | 29.0% | estimated ± 9.5 pp, medium confidence |
| DeepSeek V3 0324 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| DeepSeek V3.1 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| DeepSeek V3.1 (Reasoning) | 2.0% | 30.9% | estimated ± 9.5 pp, medium confidence |
| DeepSeek V3.2 | 0.9% | 27.0% | estimated ± 9.5 pp, medium confidence |
| Exaone 4.0 1.2B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Exaone 4.0 32B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Gemini 2.5 Flash | 1.4% | 29.0% | estimated ± 9.5 pp, medium confidence |
| Gemini 3.5 Flash-Lite | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Gemini 3 Flash | 1.4% | 29.0% | estimated ± 9.5 pp, medium confidence |
| Gemini 4 Argon | 27.1% | 55.5% | estimated ± 9.5 pp, medium confidence |
| Gemma 3 27B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| GLM-4.5-Air | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| GLM-4.6 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| GLM-5.3 | 19.1% | 51.1% | estimated ± 9.5 pp, medium confidence |
| GLM-5.3-Flash | 15.4% | 48.6% | estimated ± 9.5 pp, medium confidence |
| GLM-5-Turbo | 0.3% | 23.5% | estimated ± 9.5 pp, medium confidence |
| GLM-5V-Turbo | 0.6% | 25.5% | estimated ± 9.5 pp, medium confidence |
| GPT-4o | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| GPT-5.1-Codex | 5.7% | 38.5% | estimated ± 9.5 pp, medium confidence |
| GPT-5.1-Codex-Max | 5.7% | 38.5% | estimated ± 9.5 pp, medium confidence |
| GPT-5.2-Codex | 8.7% | 42.4% | estimated ± 9.5 pp, medium confidence |
| GPT-5.3 Codex | 16.9% | 49.6% | estimated ± 9.5 pp, medium confidence |
| GPT-5 (high) | 5.7% | 38.5% | estimated ± 9.5 pp, medium confidence |
| GPT-5 (medium) | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| GPT-OSS 120B | 1.1% | 27.8% | estimated ± 9.5 pp, medium confidence |
| GPT-OSS 20B | 1.4% | 29.0% | estimated ± 9.5 pp, medium confidence |
| Granite-4.0-350M | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Granite-4.0-H-1B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Granite-4.0-H-350M | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Grok 4 | 2.0% | 30.9% | estimated ± 9.5 pp, medium confidence |
| Grok 4.1 Fast | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Grok 4.1 Fast (Reasoning) | 2.9% | 33.2% | estimated ± 9.5 pp, medium confidence |
| Grok 4.7 | 17.7% | 50.2% | estimated ± 9.5 pp, medium confidence |
| Grok 4 Fast (Reasoning) | 2.9% | 33.2% | estimated ± 9.5 pp, medium confidence |
| Grok Code Fast 1 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Hy3 | 4.9% | 37.2% | estimated ± 9.5 pp, medium confidence |
| K-Exaone | 1.1% | 27.8% | estimated ± 9.5 pp, medium confidence |
| Kimi K2 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Kimi K2.7 Code | 10.0% | 43.8% | estimated ± 9.5 pp, medium confidence |
| LFM2.5-2.6B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| LFM2.5-8B-A1B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| LFM2.5-VL-1.6B-Extract | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Ling 3.0 Flash VL | 2.0% | 30.9% | estimated ± 9.5 pp, medium confidence |
| Ling 3.0 Tiny | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Ling 3.1 Flash | 18.0% | 50.4% | estimated ± 9.5 pp, medium confidence |
| Llama 3.1 405B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Llama 4 Maverick | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Llama 4 Scout | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| MiMo-V2.6-Flash | 12.0% | 45.8% | estimated ± 9.5 pp, medium confidence |
| MiMo-V2.6-Pro | 26.6% | 55.2% | estimated ± 9.5 pp, medium confidence |
| MiMo-V2-Omni | 1.1% | 27.8% | estimated ± 9.5 pp, medium confidence |
| MiMo-V2-Pro | 0.3% | 23.5% | estimated ± 9.5 pp, medium confidence |
| MiniMax M3 | 3.7% | 35.0% | estimated ± 9.5 pp, medium confidence |
| Mistral Large 2 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Mistral Large 3 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Mistral Large 4 | 10.6% | 44.4% | estimated ± 9.5 pp, medium confidence |
| Mistral Medium 3 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Mistral Medium 3.5 128B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Mistral Small 4 | 0.3% | 23.5% | estimated ± 9.5 pp, medium confidence |
| Mistral Small 4 (Reasoning) | 0.3% | 23.5% | estimated ± 9.5 pp, medium confidence |
| Muse Glimmer 30B | 2.6% | 32.5% | estimated ± 9.5 pp, medium confidence |
| Muse Spark 1.2 | 17.7% | 50.2% | estimated ± 9.5 pp, medium confidence |
| Muse Spark 1.3 | 24.9% | 54.4% | estimated ± 9.5 pp, medium confidence |
| Nemotron 3 Nano 30B | 0.9% | 27.0% | estimated ± 9.5 pp, medium confidence |
| Nemotron 3 Super 100B | 3.1% | 33.7% | estimated ± 9.5 pp, medium confidence |
| Nemotron Ultra 253B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| North Mini Code | 0.3% | 23.5% | estimated ± 9.5 pp, medium confidence |
| Nova Pro | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| o3 | 1.1% | 27.8% | estimated ± 9.5 pp, medium confidence |
| Phi-4 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Quasar 438B | 9.4% | 43.2% | estimated ± 9.5 pp, medium confidence |
| Qwen3.5 397B (Reasoning) | 0.9% | 27.0% | estimated ± 9.5 pp, medium confidence |
| Qwen3.8 Max Preview | 17.7% | 50.2% | estimated ± 9.5 pp, medium confidence |
| Qwen3 Max | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Qwen3-Omni-30B-A3B-Instruct | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Qwen3-Omni-30B-A3B-Thinking | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Sarvam 105B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Sarvam 30B | 0.3% | 23.5% | estimated ± 9.5 pp, medium confidence |
| Solar Pro 2 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Solar Pro 3 | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |
| Step 3.7 Flash | 2.3% | 31.7% | estimated ± 9.5 pp, medium confidence |
| Ultravox v0.6 Llama 3.3 70B | 0.0% | 19.0% | estimated ± 9.5 pp, medium confidence |