# SWE-Rebench → SWE-bench Verified

Calibration · https://benchgap.net/calibration/1487 · data retrieved 2026-10-29

SWE-bench Verified is estimated from SWE-Rebench with a Michaelis–Menten curve fitted on 8 models measured on both: y = 2.0000·x / (0.95233 + x), R² = 0.32, cross-validated error 4.8 pp. It is used for 0 estimates.

## Fitted on

| Model | SWE-Rebench | SWE-bench Verified |
|---|---|---|
| [Claude Opus 4.6](https://benchgap.net/model/claude-opus-4-6.md) | 65.3% | 80.8% |
| [Claude Sonnet 4.6](https://benchgap.net/model/claude-sonnet-4-6.md) | 60.7% | 79.6% |
| [GLM-4.7](https://benchgap.net/model/glm-4-7.md) | 58.7% | 73.8% |
| [GLM-5](https://benchgap.net/model/glm-5.md) | 62.8% | 77.8% |
| [GPT-5.3 Codex](https://benchgap.net/model/gpt-5-3-codex.md) | 58.2% | 85.0% |
| [Kimi K2.5](https://benchgap.net/model/kimi-k2-5.md) | 58.5% | 76.8% |
| [Qwen3.5-27B](https://benchgap.net/model/qwen3-5-27b.md) | 58.9% | 72.4% |
| [Qwen3.5-35B-A3B](https://benchgap.net/model/qwen3-5-35b-a3b.md) | 53.7% | 69.2% |