Calibration
Artificial Analysis Intelligence Index → HealthBench (length-adjusted)
HealthBench (length-adjusted) is estimated from Artificial Analysis Intelligence Index with a offset logistic curve fitted on 7 models measured on both: y = 0.5390 + (0.6296 − 0.5390) / (1 + exp(−73.67·(x − 0.5188))), R² = 0.81, cross-validated error 3.7 pp. It is used for 161 estimates.
| Estimated model | Artificial Analysis Intelligence Index | HealthBench (length-adjusted) | Source |
|---|---|---|---|
| A.X K2 | 21.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| Apodex 1.1 | 26.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Apodex 1.1 Mini | 26.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Celeris-1 | 6.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude 3 Haiku | 5.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude 3 Opus | 8.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude 4.1 Opus | 18.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude 4.1 Opus Thinking | 22.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude 4 Sonnet | 16.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude Haiku 5.5 | 43.4% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Claude Opus 4.5 | 23.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude Opus 4.6 | 26.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude Opus 4.7 | 30.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Claude Sonnet 5 | 38.2% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Command A+ | 22.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek-R1 | 13.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek R1 Distill Qwen 32B | 8.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek V3 | 8.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek V3 0324 | 9.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek V3.1 | 13.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek V3.1 (Reasoning) | 13.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek V3.2 | 16.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| DeepSeek V4.1 Flash | 39.5% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Exaone 4.0 1.2B | 5.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Exaone 4.0 32B | 6.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemini 1.0 Pro | 5.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemini 1.5 Pro | 7.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemini 2.5 Flash | 9.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemini 2.5 Pro | 16.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemini 3.5 Flash-Lite | 22.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemini 3 Flash | 17.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemini 4 Argon | 52.6% | 59.5% | estimated ± 3.7 pp, medium confidence |
| Gemma 3 27B | 4.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemma 4 12B | 14.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemma 4 26B A4B | 16.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemma 4 31B | 14.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemma 4 E2B | 7.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Gemma 4 E4B | 8.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-4.5-Air | 11.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-4.6 | 14.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-4.7 | 22.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-5 | 27.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-5.1 | 26.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-5.2 | 33.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-5.3 | 44.8% | 54.0% | estimated ± 3.7 pp, medium confidence |
| GLM-5.3-Flash | 41.8% | 53.9% | estimated ± 3.7 pp, medium confidence |
| GLM-5-Turbo | 26.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| GLM-5V-Turbo | 23.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-4.1 | 12.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-4.1 mini | 10.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-4.1 nano | 7.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-4 Turbo | 7.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-4o | 8.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-4o mini | 6.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-5.1-Codex | 23.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-5.1-Codex-Max | 23.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-5.2-Codex | 28.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-5.3 Codex | 32.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-5 (high) | 23.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-5 (medium) | 22.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-OSS 120B | 11.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| GPT-OSS 20B | 9.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| Granite-4.0-350M | 4.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Granite-4.0-H-1B | 5.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Granite-4.0-H-350M | 4.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Granite 4.2 30B | 14.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Granite 4.2 3B | 9.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Granite 4.2 8B | 11.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Grok 4 | 22.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| Grok 4.1 Fast | 11.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Grok 4.1 Fast (Reasoning) | 20.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Grok 4.3 | 37.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Grok 4 Fast (Reasoning) | 17.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Grok Code Fast 1 | 14.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Hy3 | 25.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Hy3 Preview | 41.2% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Inkling | 25.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| K-Exaone | 14.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| K-EXAONE 2.0 | 19.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Kimi K2.6 | 27.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| Kimi K2 | 12.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Kimi K2.5 | 23.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| Kimi K2.5 (Reasoning) | 23.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| Kimi K2.7 Code | 25.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| LFM2.5-2.6B | 8.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| LFM2.5-8B-A1B | 7.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| LFM2.5-VL-1.6B-Extract | 4.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Ling 2.6 Flash | 14.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Ling 3.0 Flash | 20.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Ling 3.0 Flash FP8 | 20.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Ling 3.0 Flash VL | 24.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Ling 3.0 Tiny | 11.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Llama 3.1 405B | 7.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Llama 4 Maverick | 10.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| Llama 4 Scout | 8.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| Mercury 2.5 | 12.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiMo-V2.5-Pro | 26.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiMo-V2.6-Flash | 37.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiMo-V2.6-Pro | 46.3% | 54.1% | estimated ± 3.7 pp, medium confidence |
| MiMo-V2-Flash | 16.1% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiMo-V2-Omni | 23.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiMo-V2-Pro | 28.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiniCPM5-2B | 12.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiniMax M2.7 | 22.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| MiniMax M3 | 29.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Mistral Large 2 | 7.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Mistral Large 3 | 9.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Mistral Large 4 | 38.4% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Mistral Medium 3 | 9.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| Mistral Medium 3.5 128B | 14.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Mistral Small 4 | 11.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Mistral Small 4 (Reasoning) | 11.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Muse Glimmer 30B | 17.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| Muse Spark | 31.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Muse Spark 1.2 | 39.6% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Muse Spark 1.3 | 48.1% | 54.4% | estimated ± 3.7 pp, medium confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 12.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Nemotron 3 Nano 30B | 8.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Nemotron 3 Nano Omni 30B A3B | 10.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Nemotron 3 Super 100B | 12.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Nemotron 3 Ultra | 22.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Nemotron Ultra 253B | 7.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| North Mini Code | 9.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Nova Pro | 7.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| o1 | 15.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| o1-preview | 11.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| o1-pro | 12.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| o3 | 20.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| o3-mini | 12.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| o3-pro | 21.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Phi-4 | 5.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Phi-4 Multimodal Instruct | 5.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Quasar 438B | 26.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen2.5 Coder 32B Instruct | 6.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.5-122B-A10B | 15.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.5-27B | 22.9% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.5-35B-A3B | 19.3% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.5 397B | 21.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.5 397B (Reasoning) | 21.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.6-27B | 21.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.6-35B-A3B | 18.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen 3.6 Max (preview) | 28.4% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.6 Plus | 27.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.7 Max | 29.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.7 Plus | 25.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.8-27B | 33.7% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3.8-Flash-Next | 39.8% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Qwen3.8 Max Preview | 45.4% | 54.0% | estimated ± 3.7 pp, medium confidence |
| Qwen3 Max | 15.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3-Omni-30B-A3B-Instruct | 6.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| Qwen3-Omni-30B-A3B-Thinking | 7.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Sarvam 105B | 8.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Sarvam 30B | 6.6% | 53.9% | estimated ± 3.7 pp, low confidence |
| Solar Pro 2 | 7.0% | 53.9% | estimated ± 3.7 pp, low confidence |
| Solar Pro 3 | 7.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Solar Pro 4 | 28.2% | 53.9% | estimated ± 3.7 pp, low confidence |
| Step 3.7 Flash | 19.5% | 53.9% | estimated ± 3.7 pp, low confidence |
| Step 5 Preview | 43.7% | 53.9% | estimated ± 3.7 pp, medium confidence |
| Trinity-Large-Preview | 10.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Trinity-Large-Thinking | 10.8% | 53.9% | estimated ± 3.7 pp, low confidence |
| Ultravox v0.6 Llama 3.3 70B | 7.7% | 53.9% | estimated ± 3.7 pp, low confidence |