Calibration
SWE-bench Pro → FrontierCode 2.0
FrontierCode 2.0 is estimated from SWE-bench Pro with a Hill curve fitted on 7 models measured on both: y = 0.1700 + (1.2000 − 0.1700)·x^6.00 / (1.06422^6.00 + x^6.00), R² = 0.69, cross-validated error 7.1 pp. It is used for 62 estimates.
| Estimated model | SWE-bench Pro | FrontierCode 2.0 | Source |
|---|---|---|---|
| Claude Mythos 5 | 80.3% | 33.0% | estimated ± 7.1 pp, low confidence |
| Sakana Fugu-Ultra | 73.7% | 27.2% | estimated ± 7.1 pp, low confidence |
| Ling 3.1 Flash | 68.7% | 24.0% | estimated ± 7.1 pp, low confidence |
| Beam 1.0 501B-A23B | 66.1% | 22.6% | estimated ± 7.1 pp, low confidence |
| Beam 1.0 501B-A23B FP8 | 65.9% | 22.5% | estimated ± 7.1 pp, low confidence |
| Hy4 preview | 65.7% | 22.4% | estimated ± 7.1 pp, low confidence |
| Beam 1.0 501B-A23B NVFP4 | 65.6% | 22.4% | estimated ± 7.1 pp, low confidence |
| Ornith-1.5-397B | 65.1% | 22.1% | estimated ± 7.1 pp, low confidence |
| Claude Opus 4.7 (Adaptive) | 64.3% | 21.8% | estimated ± 7.1 pp, low confidence |
| Qwen3.8-Omni-Flash | 63.3% | 21.4% | estimated ± 7.1 pp, low confidence |
| Qwen3.8-Flash-Next | 62.5% | 21.1% | estimated ± 7.1 pp, low confidence |
| Ornith-1.0-397B | 62.2% | 21.0% | estimated ± 7.1 pp, low confidence |
| Muse Spark 1.1 | 61.5% | 20.7% | estimated ± 7.1 pp, low confidence |
| Sage-A1.5 | 61.2% | 20.6% | estimated ± 7.1 pp, low confidence |
| dots3-note Preview | 61.0% | 20.5% | estimated ± 7.1 pp, low confidence |
| Qwen3.7 Max | 60.6% | 20.4% | estimated ± 7.1 pp, low confidence |
| Atria Dawn Preview | 59.6% | 20.1% | estimated ± 7.1 pp, low confidence |
| Ornith-1.5-35B-A3B | 59.6% | 20.1% | estimated ± 7.1 pp, low confidence |
| Laguna S 2.1 | 59.4% | 20.0% | estimated ± 7.1 pp, low confidence |
| MiniMax M3 | 59.0% | 19.9% | estimated ± 7.1 pp, low confidence |
| Sakana Fugu | 59.0% | 19.9% | estimated ± 7.1 pp, low confidence |
| Kimi K2.6 | 58.6% | 19.8% | estimated ± 7.1 pp, low confidence |
| GLM-5.1 | 58.4% | 19.7% | estimated ± 7.1 pp, low confidence |
| GPT-5.4 | 57.7% | 19.6% | estimated ± 7.1 pp, low confidence |
| Qwen3.7 Plus | 57.6% | 19.5% | estimated ± 7.1 pp, low confidence |
| Qwen 3.6 Max (preview) | 57.3% | 19.5% | estimated ± 7.1 pp, low confidence |
| MiMo-V2.5-Pro | 57.2% | 19.4% | estimated ± 7.1 pp, low confidence |
| Claude Opus 4.5 | 57.1% | 19.4% | estimated ± 7.1 pp, low confidence |
| GPT-5.3 Codex | 56.8% | 19.3% | estimated ± 7.1 pp, low confidence |
| Ling 3.0 Flash | 56.6% | 19.3% | estimated ± 7.1 pp, low confidence |
| Qwen3.6 Plus | 56.6% | 19.3% | estimated ± 7.1 pp, low confidence |
| Step 3.7 Flash | 56.3% | 19.2% | estimated ± 7.1 pp, low confidence |
| MiniMax M2.7 | 56.2% | 19.2% | estimated ± 7.1 pp, low confidence |
| MiMo-V2.5 | 56.1% | 19.2% | estimated ± 7.1 pp, low confidence |
| Inkling-Small | 55.9% | 19.1% | estimated ± 7.1 pp, low confidence |
| GPT-5.2 | 55.6% | 19.1% | estimated ± 7.1 pp, low confidence |
| DeepSeek V4 Pro 0813 | 55.4% | 19.0% | estimated ± 7.1 pp, low confidence |
| GLM-5 | 55.1% | 18.9% | estimated ± 7.1 pp, low confidence |
| Spark-X2.6-4B | 54.3% | 18.8% | estimated ± 7.1 pp, low confidence |
| Gemini 3.5 Flash-Lite | 54.2% | 18.8% | estimated ± 7.1 pp, low confidence |
| Qwen3.6-27B | 53.5% | 18.6% | estimated ± 7.1 pp, low confidence |
| Claude Opus 4.6 | 53.4% | 18.6% | estimated ± 7.1 pp, low confidence |
| MAI-Thinking-1 | 52.8% | 18.5% | estimated ± 7.1 pp, low confidence |
| DeepSeek V4 Flash 0731 | 52.6% | 18.5% | estimated ± 7.1 pp, low confidence |
| Muse Spark | 52.4% | 18.5% | estimated ± 7.1 pp, low confidence |
| Grok 4.20 | 51.8% | 18.4% | estimated ± 7.1 pp, low confidence |
| Nemotron 3.5 Super VL 120B A12B | 51.8% | 18.4% | estimated ± 7.1 pp, low confidence |
| Nemotron 3.5 Super VL 120B A12B NVFP4 | 51.3% | 18.3% | estimated ± 7.1 pp, low confidence |
| Muse Glimmer 30B | 51.2% | 18.3% | estimated ± 7.1 pp, low confidence |
| Qwen3.5 397B | 50.9% | 18.2% | estimated ± 7.1 pp, low confidence |
| Kimi K2.5 | 50.7% | 18.2% | estimated ± 7.1 pp, low confidence |
| Ornith-1.0-35B | 50.4% | 18.2% | estimated ± 7.1 pp, low confidence |
| Qwen3.6-35B-A3B | 49.5% | 18.0% | estimated ± 7.1 pp, low confidence |
| Laguna M.1 | 49.2% | 18.0% | estimated ± 7.1 pp, low confidence |
| Laguna XS 2.1 | 47.6% | 17.8% | estimated ± 7.1 pp, low confidence |
| Ornith-1.5-9B | 47.5% | 17.8% | estimated ± 7.1 pp, low confidence |
| Laguna XS.2 | 46.3% | 17.7% | estimated ± 7.1 pp, low confidence |
| Ornith-1.0-9B | 42.9% | 17.4% | estimated ± 7.1 pp, low confidence |
| LongCat-Flash-Lite-Sparse | 40.6% | 17.3% | estimated ± 7.1 pp, low confidence |
| Granite 4.2 30B | 33.3% | 17.1% | estimated ± 7.1 pp, low confidence |
| LLaDA2.2-flash | 30.1% | 17.1% | estimated ± 7.1 pp, low confidence |
| Mellum2.1-12B-A2.5B-Thinking | 28.0% | 17.0% | estimated ± 7.1 pp, low confidence |
Fitted on
| Model | SWE-bench Pro | FrontierCode 2.0 |
|---|---|---|
| Claude Fable 5.1 | 81.2% | 32.4% |
| Claude Opus 4.8 | 69.2% | 14.7% |
| Claude Opus 5 | 79.2% | 26.7% |
| Claude Opus 5.5 | 89.9% | 45.5% |
| Claude Sonnet 5.5 | 81.3% | 41.0% |
| GPT-5.5 | 58.6% | 22.6% |
| GPT-5.6 Sol | 64.6% | 27.6% |