Benchmark
SWE-bench Pro
Harness and split version must be recorded with any score.
- Released
- 2025-09-23
- Categories
- coding_agent
- Openness
- unknown
- Source
- model_reports
- Reported scores
- 10
Reported scores
model_reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Claude Opus 5 | Anthropic | 79.2 | 2026-07-24 | evidence |
| DeepSeek-V4-Flash | DeepSeek | 52.6 | 2026-06-25 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 52.1 | 2026-04-22 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 54.4 | 2026-04-22 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 55.4 | 2026-04-22 | evidence |
| Gemini 3.1 Pro | 54.2 | 2026-02-19 | evidence | |
| GLM-5.1 | Z.ai | 58.4 | 2026-04-03 | evidence |
| GLM-5.2 | Z.ai | 62.1 | 2026-06-16 | evidence |
| Grok 4.5 | xAI | 64.7 | 2026-07-08 | evidence |
| Hy4 preview | Tencent | 65.7 | 2026-08-28 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.