Calibration
Vals SWE-bench → DeepSWE
DeepSWE is estimated from Vals SWE-bench with a offset logistic curve fitted on 18 models measured on both: y = 0.5632 + (0.6782 − 0.5632) / (1 + exp(−200.00·(x − 0.9175))), R² = 0.43, cross-validated error 7.2 pp. It is used for 21 estimates.
| Estimated model | Vals SWE-bench | DeepSWE | Source |
|---|---|---|---|
| Inkling-Small | 82.2% | 56.3% | estimated ± 7.2 pp, low confidence |
| Composer 2.5 | 79.6% | 56.3% | estimated ± 7.2 pp, low confidence |
| GPT-5.3 Codex | 78.0% | 56.3% | estimated ± 7.2 pp, low confidence |
| Claude Sonnet 4.6 | 77.4% | 56.3% | estimated ± 7.2 pp, low confidence |
| Gemini 3.5 Flash-Lite | 75.0% | 56.3% | estimated ± 7.2 pp, low confidence |
| Gemini 3 Flash | 75.0% | 56.3% | estimated ± 7.2 pp, low confidence |
| Claude Haiku 4.5 | 66.6% | 56.3% | estimated ± 7.2 pp, low confidence |
| Gemini 3.1 Flash-Lite | 62.8% | 56.3% | estimated ± 7.2 pp, low confidence |
| GLM-4.7 | 69.4% | 56.3% | estimated ± 7.2 pp, low confidence |
| GPT-5.2-Codex | 72.4% | 56.3% | estimated ± 7.2 pp, low confidence |
| Grok 4.20 | 72.2% | 56.3% | estimated ± 7.2 pp, low confidence |
| Grok 4.3 | 71.4% | 56.3% | estimated ± 7.2 pp, low confidence |
| Ling 3.0 Flash | 65.2% | 56.3% | estimated ± 7.2 pp, low confidence |
| MiMo-V2.5 | 71.0% | 56.3% | estimated ± 7.2 pp, low confidence |
| MiniMax M2.7 | 73.8% | 56.3% | estimated ± 7.2 pp, low confidence |
| Mistral Medium 3.5 128B | 66.4% | 56.3% | estimated ± 7.2 pp, low confidence |
| Nemotron 3 Ultra | 69.0% | 56.3% | estimated ± 7.2 pp, low confidence |
| Qwen3.5 Flash | 64.4% | 56.3% | estimated ± 7.2 pp, low confidence |
| Qwen3.6-27B | 70.0% | 56.3% | estimated ± 7.2 pp, low confidence |
| Qwen 3.6 Max (preview) | 72.8% | 56.3% | estimated ± 7.2 pp, low confidence |
| Qwen3.6 Plus | 73.4% | 56.3% | estimated ± 7.2 pp, low confidence |
Fitted on
| Model | Vals SWE-bench | DeepSWE |
|---|---|---|
| Claude Opus 5 | 97.0% | 68.8% |
| DeepSeek V4 Flash 0731 | 88.8% | 54.4% |
| DeepSeek V4 Pro 0813 | 96.4% | 62.7% |
| Gemini 3.6 Flash | 79.6% | 49.0% |
| Gemini 3.7 Flash | 80.8% | 65.3% |
| Gemini 3.8 Flash | 80.0% | 73.8% |
| GLM-5.3 | 95.4% | 66.9% |
| GLM-5.3-Flash | 92.0% | 63.4% |
| GPT-5.6 Luna | 93.0% | 67.2% |
| GPT-5.6 Sol | 96.2% | 72.7% |
| GPT-5.6 Terra | 95.4% | 69.6% |
| Grok 4.5 | 86.6% | 53.0% |
| Grok 4.6 | 95.6% | 65.9% |
| Kimi K3 | 93.4% | 67.5% |
| Muse Spark 1.1 | 82.0% | 53.3% |
| Muse Spark 1.2 | 86.6% | 59.3% |
| Qwen3.8-27B | 86.0% | 42.2% |
| Qwen3.8 Max | 85.6% | 56.6% |