Calibration
SWE-bench Verified → SWE-bench Pro
SWE-bench Pro is estimated from SWE-bench Verified with a linear curve fitted on 50 models measured on both: y = 1.1067·x + -0.3019, R² = 0.90, cross-validated error 4.6 pp. It is used for 61 estimates.
| Estimated model | SWE-bench Verified | SWE-bench Pro | Source |
|---|---|---|---|
| Ember-1 | 92.2% | 71.8% | estimated ± 4.6 pp, high confidence |
| BTL-4 | 78.4% | 56.6% | estimated ± 4.6 pp, high confidence |
| MiMo-V2-Pro | 78.0% | 56.1% | estimated ± 4.6 pp, high confidence |
| Apodex 1.1 | 77.7% | 55.8% | estimated ± 4.6 pp, high confidence |
| Mistral Medium 3.5 128B | 77.6% | 55.7% | estimated ± 4.6 pp, high confidence |
| Claude Sonnet 4.5 | 77.2% | 55.2% | estimated ± 4.6 pp, high confidence |
| Kimi K2.5 (Reasoning) | 76.8% | 54.8% | estimated ± 4.6 pp, high confidence |
| MiMo-V2-Omni | 74.8% | 52.6% | estimated ± 4.6 pp, high confidence |
| Claude 4.1 Opus | 74.5% | 52.3% | estimated ± 4.6 pp, high confidence |
| Hy3 Preview | 74.4% | 52.2% | estimated ± 4.6 pp, high confidence |
| Step 3.5 Flash | 74.4% | 52.2% | estimated ± 4.6 pp, high confidence |
| Flint | 74.2% | 51.9% | estimated ± 4.6 pp, high confidence |
| GLM-4.7 | 73.8% | 51.5% | estimated ± 4.6 pp, high confidence |
| MiMo-V2-Flash | 73.4% | 51.0% | estimated ± 4.6 pp, high confidence |
| Claude Haiku 4.5 | 73.3% | 50.9% | estimated ± 4.6 pp, high confidence |
| Jais-3-MoVA-36B-A4B | 72.8% | 50.4% | estimated ± 4.6 pp, high confidence |
| Claude 4 Sonnet | 72.7% | 50.3% | estimated ± 4.6 pp, high confidence |
| MAI-Code-1.1-Flash | 72.6% | 50.2% | estimated ± 4.6 pp, high confidence |
| Claude 4 Opus | 72.5% | 50.0% | estimated ± 4.6 pp, high confidence |
| Jais-3-32B | 72.4% | 49.9% | estimated ± 4.6 pp, high confidence |
| Qwen3.5-27B | 72.4% | 49.9% | estimated ± 4.6 pp, high confidence |
| Qwen3.5-122B-A10B | 72.0% | 49.5% | estimated ± 4.6 pp, high confidence |
| Nemotron 3 Ultra | 71.9% | 49.4% | estimated ± 4.6 pp, high confidence |
| Grok Code Fast 1 | 70.8% | 48.2% | estimated ± 4.6 pp, high confidence |
| Solar Pro 4 | 70.6% | 47.9% | estimated ± 4.6 pp, high confidence |
| Solar Open 2 | 70.4% | 47.7% | estimated ± 4.6 pp, high confidence |
| Jais-3-375B-A23B | 70.2% | 47.5% | estimated ± 4.6 pp, high confidence |
| Jais-3-7B | 70.2% | 47.5% | estimated ± 4.6 pp, high confidence |
| Qwen3-Coder 480B-A35B | 69.6% | 46.8% | estimated ± 4.6 pp, high confidence |
| Qwen3.5-35B-A3B | 69.2% | 46.4% | estimated ± 4.6 pp, high confidence |
| K-EXAONE 2.0 | 68.2% | 45.3% | estimated ± 4.6 pp, high confidence |
| DeepSeek V3.1 | 66.0% | 42.9% | estimated ± 4.6 pp, high confidence |
| Kimi K2 | 65.8% | 42.6% | estimated ± 4.6 pp, high confidence |
| Gemini 2.5 Pro | 63.8% | 40.4% | estimated ± 4.6 pp, high confidence |
| GPT-OSS 120B | 62.4% | 38.9% | estimated ± 4.6 pp, high confidence |
| Claude 3.7 Sonnet | 62.3% | 38.8% | estimated ± 4.6 pp, high confidence |
| Ternary Bonsai 2 27B | 60.8% | 37.1% | estimated ± 4.6 pp, high confidence |
| GPT-OSS 20B | 60.7% | 37.0% | estimated ± 4.6 pp, high confidence |
| Nemotron 3 Super 100B | 60.5% | 36.7% | estimated ± 4.6 pp, high confidence |
| GLM-4.7-Flash | 59.2% | 35.3% | estimated ± 4.6 pp, high confidence |
| DeepSeek-R1-0528 | 57.6% | 33.6% | estimated ± 4.6 pp, high confidence |
| MiniMax M1 80k | 56.0% | 31.8% | estimated ± 4.6 pp, high confidence |
| GPT-4.1 | 54.6% | 30.2% | estimated ± 4.6 pp, high confidence |
| ZAYA1-74B-Preview | 53.2% | 28.7% | estimated ± 4.6 pp, high confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 52.8% | 28.2% | estimated ± 4.6 pp, high confidence |
| LFM2.5-24B-A2B | 51.0% | 26.3% | estimated ± 4.6 pp, high confidence |
| K-Exaone | 49.4% | 24.5% | estimated ± 4.6 pp, high confidence |
| o3-mini | 49.3% | 24.4% | estimated ± 4.6 pp, high confidence |
| DeepSeek-R1 | 49.2% | 24.3% | estimated ± 4.6 pp, high confidence |
| Claude 3.5 Sonnet | 49.0% | 24.0% | estimated ± 4.6 pp, high confidence |
| Devstral Small | 46.8% | 21.6% | estimated ± 4.6 pp, high confidence |
| Sarvam 105B | 45.0% | 19.6% | estimated ± 4.6 pp, medium confidence |
| LFM3-1.7B | 43.4% | 17.8% | estimated ± 4.6 pp, medium confidence |
| DeepSeek V3 | 42.0% | 16.3% | estimated ± 4.6 pp, medium confidence |
| Claude 3.5 Haiku | 40.6% | 14.7% | estimated ± 4.6 pp, medium confidence |
| Sarvam 30B | 34.0% | 7.4% | estimated ± 4.6 pp, medium confidence |
| Gemini 2.5 Flash-Lite | 31.6% | 4.8% | estimated ± 4.6 pp, medium confidence |
| Gemini 1.5 Flash | 9.6% | 0.0% | estimated ± 4.6 pp, medium confidence |
| Gemini 2.0 Flash | 21.4% | 0.0% | estimated ± 4.6 pp, medium confidence |
| Gemini 2.0 Flash-Lite | 12.5% | 0.0% | estimated ± 4.6 pp, medium confidence |
| GPT-4.1 mini | 23.6% | 0.0% | estimated ± 4.6 pp, medium confidence |