Calibration
SWE-bench Verified → Vals LiveCodeBench
Vals LiveCodeBench is estimated from SWE-bench Verified with a Hill curve fitted on 21 models measured on both: y = 0.0000 + (1.0017 − 0.0000)·x^6.00 / (0.61912^6.00 + x^6.00), R² = 0.39, cross-validated error 9.9 pp. It is used for 34 estimates.
| Estimated model | SWE-bench Verified | Vals LiveCodeBench | Source |
|---|---|---|---|
| Ember-1 | 92.2% | 91.8% | estimated ± 9.9 pp, low confidence |
| BTL-4 | 78.4% | 80.6% | estimated ± 9.9 pp, low confidence |
| MiMo-V2-Pro | 78.0% | 80.1% | estimated ± 9.9 pp, low confidence |
| Apodex 1.1 | 77.7% | 79.8% | estimated ± 9.9 pp, low confidence |
| Mistral Medium 3.5 128B | 77.6% | 79.6% | estimated ± 9.9 pp, low confidence |
| Claude Sonnet 4.5 | 77.2% | 79.1% | estimated ± 9.9 pp, low confidence |
| MiMo-V2-Omni | 74.8% | 75.8% | estimated ± 9.9 pp, low confidence |
| Claude 4.1 Opus | 74.5% | 75.3% | estimated ± 9.9 pp, low confidence |
| Jais-3-MoVA-36B-A4B | 72.8% | 72.7% | estimated ± 9.9 pp, low confidence |
| Claude 4 Sonnet | 72.7% | 72.5% | estimated ± 9.9 pp, low confidence |
| MAI-Code-1.1-Flash | 72.6% | 72.3% | estimated ± 9.9 pp, low confidence |
| Claude 4 Opus | 72.5% | 72.2% | estimated ± 9.9 pp, low confidence |
| Jais-3-32B | 72.4% | 72.0% | estimated ± 9.9 pp, low confidence |
| Solar Pro 4 | 70.6% | 68.9% | estimated ± 9.9 pp, low confidence |
| Jais-3-375B-A23B | 70.2% | 68.1% | estimated ± 9.9 pp, low confidence |
| Jais-3-7B | 70.2% | 68.1% | estimated ± 9.9 pp, low confidence |
| Qwen3-Coder 480B-A35B | 69.6% | 67.0% | estimated ± 9.9 pp, low confidence |
| Qwen3.5-35B-A3B | 69.2% | 66.2% | estimated ± 9.9 pp, low confidence |
| DeepSeek V3.1 | 66.0% | 59.6% | estimated ± 9.9 pp, low confidence |
| Kimi K2 | 65.8% | 59.1% | estimated ± 9.9 pp, low confidence |
| Gemini 2.5 Pro | 63.8% | 54.6% | estimated ± 9.9 pp, low confidence |
| Claude 3.7 Sonnet | 62.3% | 51.0% | estimated ± 9.9 pp, low confidence |
| Nemotron 3 Super 100B | 60.5% | 46.5% | estimated ± 9.9 pp, low confidence |
| GLM-4.7-Flash | 59.2% | 43.4% | estimated ± 9.9 pp, low confidence |
| DeepSeek-R1-0528 | 57.6% | 39.4% | estimated ± 9.9 pp, low confidence |
| MiniMax M1 80k | 56.0% | 35.4% | estimated ± 9.9 pp, low confidence |
| GPT-4.1 | 54.6% | 32.0% | estimated ± 9.9 pp, low confidence |
| ZAYA1-74B-Preview | 53.2% | 28.7% | estimated ± 9.9 pp, low confidence |
| K-Exaone | 49.4% | 20.5% | estimated ± 9.9 pp, low confidence |
| o3-mini | 49.3% | 20.3% | estimated ± 9.9 pp, low confidence |
| DeepSeek-R1 | 49.2% | 20.2% | estimated ± 9.9 pp, low confidence |
| Claude 3.5 Sonnet | 49.0% | 19.8% | estimated ± 9.9 pp, low confidence |
| Devstral Small | 46.8% | 15.7% | estimated ± 9.9 pp, low confidence |
| Sarvam 105B | 45.0% | 12.9% | estimated ± 9.9 pp, low confidence |
Fitted on
| Model | SWE-bench Verified | Vals LiveCodeBench |
|---|---|---|
| Claude Fable 5 | 95.0% | 89.8% |
| Claude Haiku 4.5 | 73.3% | 41.2% |
| Claude Opus 4.8 | 88.6% | 87.8% |
| Claude Opus 5 | 96.0% | 89.0% |
| Claude Sonnet 4.6 | 79.6% | 82.1% |
| Claude Sonnet 5 | 85.2% | 82.4% |
| DeepSeek V4 Flash 0731 | 79.0% | 87.3% |
| DeepSeek V4 Pro 0813 | 80.6% | 87.5% |
| GLM-4.7 | 73.8% | 82.2% |
| GPT-5.3 Codex | 85.0% | 87.3% |
| Grok 4.20 | 76.7% | 84.3% |
| Grok Code Fast 1 | 70.8% | 62.0% |
| Inkling | 77.6% | 85.5% |
| Inkling-Small | 80.2% | 85.9% |
| Kimi K2.6 | 80.2% | 86.8% |
| Laguna M.1 | 74.6% | 68.1% |
| Laguna XS.2 | 69.9% | 67.8% |
| MiniMax M3 | 80.5% | 82.2% |
| Nemotron 3 Ultra | 71.9% | 86.0% |
| Qwen3.6 Plus | 78.8% | 86.0% |
| Qwen3.7 Max | 80.4% | 87.1% |