Calibration
Vals SWE-bench → PostTrainBench v1.1
PostTrainBench v1.1 is estimated from Vals SWE-bench with a inverse Michaelis–Menten curve fitted on 9 models measured on both: y = 0.38080·(x − 0.0000) / (0.0000 + 2.0000 − x), R² = 0.61, cross-validated error 3.3 pp. It is used for 48 estimates.
| Estimated model | Vals SWE-bench | PostTrainBench v1.1 | Source |
|---|---|---|---|
| DeepSeek V4 Pro 0813 | 96.4% | 35.4% | estimated ± 3.3 pp, high confidence |
| Grok 4.6 | 95.6% | 34.9% | estimated ± 3.3 pp, high confidence |
| GLM-5.3 | 95.4% | 34.7% | estimated ± 3.3 pp, high confidence |
| GPT-5.6 Terra | 95.4% | 34.7% | estimated ± 3.3 pp, high confidence |
| Claude Fable 5 | 95.0% | 34.5% | estimated ± 3.3 pp, high confidence |
| GPT-5.6 Luna | 93.0% | 33.1% | estimated ± 3.3 pp, high confidence |
| GLM-5.3-Flash | 92.0% | 32.4% | estimated ± 3.3 pp, high confidence |
| DeepSeek V4 Flash 0731 | 88.8% | 30.4% | estimated ± 3.3 pp, high confidence |
| Muse Spark 1.2 | 86.6% | 29.1% | estimated ± 3.3 pp, high confidence |
| Qwen3.8-27B | 86.0% | 28.7% | estimated ± 3.3 pp, high confidence |
| Qwen3.8 Max | 85.6% | 28.5% | estimated ± 3.3 pp, high confidence |
| Inkling-Small | 82.2% | 26.6% | estimated ± 3.3 pp, high confidence |
| Muse Spark 1.1 | 82.0% | 26.5% | estimated ± 3.3 pp, high confidence |
| Gemini 3.7 Flash | 80.8% | 25.8% | estimated ± 3.3 pp, high confidence |
| Gemini 3.8 Flash | 80.0% | 25.4% | estimated ± 3.3 pp, high confidence |
| Claude Sonnet 5 | 79.6% | 25.2% | estimated ± 3.3 pp, high confidence |
| Composer 2.5 | 79.6% | 25.2% | estimated ± 3.3 pp, high confidence |
| Gemini 3.6 Flash | 79.6% | 25.2% | estimated ± 3.3 pp, high confidence |
| Gemini 3.5 Flash | 78.8% | 24.8% | estimated ± 3.3 pp, high confidence |
| Kimi K2.7 Code | 78.2% | 24.4% | estimated ± 3.3 pp, medium confidence |
| GPT-5.3 Codex | 78.0% | 24.3% | estimated ± 3.3 pp, medium confidence |
| Inkling | 77.6% | 24.1% | estimated ± 3.3 pp, medium confidence |
| Claude Sonnet 4.6 | 77.4% | 24.0% | estimated ± 3.3 pp, medium confidence |
| GLM-5.1 | 76.4% | 23.5% | estimated ± 3.3 pp, medium confidence |
| Kimi K2.6 | 76.2% | 23.4% | estimated ± 3.3 pp, medium confidence |
| Gemini 3.5 Flash-Lite | 75.0% | 22.8% | estimated ± 3.3 pp, medium confidence |
| Gemini 3 Flash | 75.0% | 22.8% | estimated ± 3.3 pp, medium confidence |
| MiniMax M3 | 75.0% | 22.8% | estimated ± 3.3 pp, medium confidence |
| Muse Spark | 74.4% | 22.6% | estimated ± 3.3 pp, medium confidence |
| MiMo-V2.5-Pro | 74.0% | 22.4% | estimated ± 3.3 pp, medium confidence |
| MiniMax M2.7 | 73.8% | 22.3% | estimated ± 3.3 pp, medium confidence |
| Qwen3.6 Plus | 73.4% | 22.1% | estimated ± 3.3 pp, medium confidence |
| GPT-5.4 mini | 73.0% | 21.9% | estimated ± 3.3 pp, medium confidence |
| Qwen 3.6 Max (preview) | 72.8% | 21.8% | estimated ± 3.3 pp, medium confidence |
| GPT-5.2-Codex | 72.4% | 21.6% | estimated ± 3.3 pp, medium confidence |
| Grok 4.20 | 72.2% | 21.5% | estimated ± 3.3 pp, medium confidence |
| Grok 4.3 | 71.4% | 21.1% | estimated ± 3.3 pp, medium confidence |
| MiMo-V2.5 | 71.0% | 21.0% | estimated ± 3.3 pp, medium confidence |
| Qwen3.6-27B | 70.0% | 20.5% | estimated ± 3.3 pp, medium confidence |
| GPT-5.4 nano | 69.8% | 20.4% | estimated ± 3.3 pp, medium confidence |
| GLM-4.7 | 69.4% | 20.2% | estimated ± 3.3 pp, medium confidence |
| Nemotron 3 Ultra | 69.0% | 20.1% | estimated ± 3.3 pp, medium confidence |
| Qwen3.7 Max | 68.8% | 20.0% | estimated ± 3.3 pp, medium confidence |
| Claude Haiku 4.5 | 66.6% | 19.0% | estimated ± 3.3 pp, medium confidence |
| Mistral Medium 3.5 128B | 66.4% | 18.9% | estimated ± 3.3 pp, medium confidence |
| Ling 3.0 Flash | 65.2% | 18.4% | estimated ± 3.3 pp, medium confidence |
| Qwen3.5 Flash | 64.4% | 18.1% | estimated ± 3.3 pp, medium confidence |
| Gemini 3.1 Flash-Lite | 62.8% | 17.4% | estimated ± 3.3 pp, medium confidence |
Fitted on
| Model | Vals SWE-bench | PostTrainBench v1.1 |
|---|---|---|
| Claude Opus 4.7 | 82.0% | 28.6% |
| Claude Opus 4.8 | 88.6% | 32.9% |
| Claude Opus 5 | 97.0% | 35.0% |
| Gemini 3.1 Pro | 78.8% | 22.0% |
| GLM-5.2 | 82.8% | 31.7% |
| GPT-5.5 | 82.6% | 27.2% |
| GPT-5.6 Sol | 96.2% | 36.2% |
| Grok 4.5 | 86.6% | 23.5% |
| Kimi K3 | 93.4% | 32.0% |