Calibration
SWE-bench Verified → SciCode
SciCode is estimated from SWE-bench Verified with a Michaelis–Menten curve fitted on 19 models measured on both: y = 2.0000·x / (2.53169 + x), R² = 0.85, cross-validated error 3.3 pp. It is used for 67 estimates.
| Estimated model | SWE-bench Verified | SciCode | Source |
|---|---|---|---|
| Claude Mythos 5 | 95.5% | 54.8% | estimated ± 3.3 pp, medium confidence |
| Ember-1 | 92.2% | 53.4% | estimated ± 3.3 pp, medium confidence |
| Claude Opus 4.7 (Adaptive) | 87.6% | 51.4% | estimated ± 3.3 pp, medium confidence |
| Ornith-1.5-397B | 86.0% | 50.7% | estimated ± 3.3 pp, medium confidence |
| Ornith-1.0-397B | 82.4% | 49.1% | estimated ± 3.3 pp, medium confidence |
| Claude Opus 4.5 | 80.9% | 48.4% | estimated ± 3.3 pp, high confidence |
| Claude Opus 4.6 | 80.8% | 48.4% | estimated ± 3.3 pp, high confidence |
| GPT-5.2 | 80.0% | 48.0% | estimated ± 3.3 pp, high confidence |
| Ornith-1.5-35B-A3B | 79.0% | 47.6% | estimated ± 3.3 pp, high confidence |
| BTL-4 | 78.4% | 47.3% | estimated ± 3.3 pp, high confidence |
| dots3-note Preview | 78.4% | 47.3% | estimated ± 3.3 pp, high confidence |
| MiMo-V2-Pro | 78.0% | 47.1% | estimated ± 3.3 pp, high confidence |
| GLM-5 | 77.8% | 47.0% | estimated ± 3.3 pp, high confidence |
| Muse Spark | 77.4% | 46.8% | estimated ± 3.3 pp, high confidence |
| Claude Sonnet 4.5 | 77.2% | 46.7% | estimated ± 3.3 pp, high confidence |
| Kimi K2.5 (Reasoning) | 76.8% | 46.5% | estimated ± 3.3 pp, high confidence |
| Ornith-1.0-35B | 75.6% | 46.0% | estimated ± 3.3 pp, high confidence |
| MiMo-V2-Omni | 74.8% | 45.6% | estimated ± 3.3 pp, high confidence |
| Laguna M.1 | 74.6% | 45.5% | estimated ± 3.3 pp, high confidence |
| Claude 4.1 Opus | 74.5% | 45.5% | estimated ± 3.3 pp, high confidence |
| Step 3.5 Flash | 74.4% | 45.4% | estimated ± 3.3 pp, high confidence |
| Flint | 74.2% | 45.3% | estimated ± 3.3 pp, high confidence |
| MAI-Thinking-1 | 73.5% | 45.0% | estimated ± 3.3 pp, high confidence |
| MiMo-V2-Flash | 73.4% | 45.0% | estimated ± 3.3 pp, high confidence |
| Claude Haiku 4.5 | 73.3% | 44.9% | estimated ± 3.3 pp, high confidence |
| Jais-3-MoVA-36B-A4B | 72.8% | 44.7% | estimated ± 3.3 pp, high confidence |
| Claude 4 Sonnet | 72.7% | 44.6% | estimated ± 3.3 pp, high confidence |
| MAI-Code-1.1-Flash | 72.6% | 44.6% | estimated ± 3.3 pp, high confidence |
| Claude 4 Opus | 72.5% | 44.5% | estimated ± 3.3 pp, high confidence |
| Jais-3-32B | 72.4% | 44.5% | estimated ± 3.3 pp, high confidence |
| Qwen3.5-27B | 72.4% | 44.5% | estimated ± 3.3 pp, high confidence |
| Laguna XS 2.1 | 70.9% | 43.8% | estimated ± 3.3 pp, high confidence |
| Grok Code Fast 1 | 70.8% | 43.7% | estimated ± 3.3 pp, high confidence |
| Ornith-1.5-9B | 70.6% | 43.6% | estimated ± 3.3 pp, high confidence |
| Solar Open 2 | 70.4% | 43.5% | estimated ± 3.3 pp, high confidence |
| Jais-3-375B-A23B | 70.2% | 43.4% | estimated ± 3.3 pp, high confidence |
| Jais-3-7B | 70.2% | 43.4% | estimated ± 3.3 pp, high confidence |
| Laguna XS.2 | 69.9% | 43.3% | estimated ± 3.3 pp, high confidence |
| Qwen3-Coder 480B-A35B | 69.6% | 43.1% | estimated ± 3.3 pp, high confidence |
| Ornith-1.0-9B | 69.4% | 43.0% | estimated ± 3.3 pp, high confidence |
| Qwen3.5-35B-A3B | 69.2% | 42.9% | estimated ± 3.3 pp, high confidence |
| LongCat-Flash-Lite-Sparse | 68.2% | 42.4% | estimated ± 3.3 pp, high confidence |
| DeepSeek V3.1 | 66.0% | 41.4% | estimated ± 3.3 pp, high confidence |
| Kimi K2 | 65.8% | 41.3% | estimated ± 3.3 pp, high confidence |
| Claude 3.7 Sonnet | 62.3% | 39.5% | estimated ± 3.3 pp, high confidence |
| Ternary Bonsai 2 27B | 60.8% | 38.7% | estimated ± 3.3 pp, high confidence |
| Nemotron 3 Super 100B | 60.5% | 38.6% | estimated ± 3.3 pp, high confidence |
| GLM-4.7-Flash | 59.2% | 37.9% | estimated ± 3.3 pp, high confidence |
| DeepSeek-R1-0528 | 57.6% | 37.1% | estimated ± 3.3 pp, high confidence |
| MiniMax M1 80k | 56.0% | 36.2% | estimated ± 3.3 pp, high confidence |
| GPT-4.1 | 54.6% | 35.5% | estimated ± 3.3 pp, high confidence |
| ZAYA1-74B-Preview | 53.2% | 34.7% | estimated ± 3.3 pp, high confidence |
| K-Exaone | 49.4% | 32.7% | estimated ± 3.3 pp, high confidence |
| o3-mini | 49.3% | 32.6% | estimated ± 3.3 pp, high confidence |
| LLaDA2.2-flash | 49.3% | 32.6% | estimated ± 3.3 pp, high confidence |
| DeepSeek-R1 | 49.2% | 32.5% | estimated ± 3.3 pp, high confidence |
| Claude 3.5 Sonnet | 49.0% | 32.4% | estimated ± 3.3 pp, high confidence |
| Mellum2.1-12B-A2.5B-Thinking | 47.0% | 31.3% | estimated ± 3.3 pp, high confidence |
| Devstral Small | 46.8% | 31.2% | estimated ± 3.3 pp, high confidence |
| Sarvam 105B | 45.0% | 30.2% | estimated ± 3.3 pp, high confidence |
| Claude 3.5 Haiku | 40.6% | 27.6% | estimated ± 3.3 pp, medium confidence |
| Sarvam 30B | 34.0% | 23.7% | estimated ± 3.3 pp, medium confidence |
| Gemini 2.5 Flash-Lite | 31.6% | 22.2% | estimated ± 3.3 pp, medium confidence |
| GPT-4.1 mini | 23.6% | 17.1% | estimated ± 3.3 pp, medium confidence |
| Gemini 2.0 Flash | 21.4% | 15.6% | estimated ± 3.3 pp, medium confidence |
| Gemini 2.0 Flash-Lite | 12.5% | 9.4% | estimated ± 3.3 pp, medium confidence |
| Gemini 1.5 Flash | 9.6% | 7.3% | estimated ± 3.3 pp, medium confidence |
Fitted on
| Model | SWE-bench Verified | SciCode |
|---|---|---|
| Granite 4.2 30B | 57.0% | 38.8% |
| Granite 4.2 8B | 47.7% | 36.1% |
| Hy3 Preview | 74.4% | 41.2% |
| Inkling-Small | 80.2% | 48.7% |
| K-EXAONE 2.0 | 68.2% | 37.4% |
| Kimi K2.6 | 80.2% | 52.2% |
| Kimi K2.5 | 76.8% | 48.7% |
| LFM2.5-24B-A2B | 51.0% | 32.7% |
| LFM3-1.7B | 43.4% | 28.5% |
| MiniCPM5-2B | 46.4% | 26.3% |
| Muse Glimmer 30B | 76.0% | 43.6% |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 52.8% | 31.4% |
| Nemotron 3 Ultra | 71.9% | 44.6% |
| Qwen3.7 Max | 80.4% | 53.5% |
| Qwen3.7 Plus | 77.7% | 51.3% |
| Beam 1.0 501B-A23B | 81.6% | 49.2% |
| Beam 1.0 501B-A23B FP8 | 80.2% | 48.3% |
| Beam 1.0 501B-A23B NVFP4 | 80.7% | 46.9% |
| Spark-X2.6-4B | 62.8% | 36.8% |