Benchmark
SWE-rebench 2026-03
Coding benchmarks compare models on programming tasks. Scores are percentages.
- Models measured
- 18
- Scores recorded
- 18
- Sources
- 1
- IQ weight
- 2.15
Score distribution
- 46.8%
- Mean
- 51.3%
- Median
- 10.8%
- Std dev
- 36.1%
- 10th percentile
- 54.3%
- 90th percentile
- 13.5%
- Min
This benchmark has a weight of 2.15 for chat models, measured over 18 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | |
|---|---|---|
| 1 | GPT 5.5tools, xhigh· SWE-rebench | 63% |
| 2 | Claude Opus 4.8tools, xhigh· SWE-rebench | 56.5% |
| 3 | Claude Sonnet 4.6tools· SWE-rebench | 54.3% |
| 4 | Gemini 3.1 Pro Previewtools· SWE-rebench | 53.9% |
| 5 | Claude Opus 4.7tools, high· SWE-rebench | 53.5% |
| 6 | Gemini 3.5 Flashtools· SWE-rebench | 53.5% |
| 7 | GLM 5.2tools, high· SWE-rebench | 53.5% |
| 8 | GLM 5.1tools· SWE-rebench | 53% |
| 9 | Claude Opus 4.6tools, high· SWE-rebench | 51.3% |
| 10 | Kimi K2.6tools· SWE-rebench | 47.4% |
| 11 | DeepSeek V4 Protools, high· SWE-rebench | 46.7% |
| 12 | MiniMax M3tools· SWE-rebench | 46.1% |
| 13 | MiMo V2.5 Protools· SWE-rebench | 43.9% |
| 14 | GLM 4.7tools· SWE-rebench | 41.7% |
| 15 | Qwen3.6 27Btools· SWE-rebench | 38.7% |
| 16 | DeepSeek V4 Flashtools, high· SWE-rebench | 36.1% |
| 17 | Qwen3.6 35B A3Btools· SWE-rebench | 36.1% |
| 18 | Gemma 4 31B ITtools· SWE-rebench | 13.5% |
Tokens per task, input and output together, as measured by the source. Comparable within this benchmark only.