SWT-bench — Distinct named SWE-bench-adjacent test-writing benchmark
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | GPT-5.2 | OpenAI | 80.7 | 1 | 2026-06-05 |
| 2 | Claude Opus 4.5 | Anthropic | 80.2 | 1 | 2026-06-05 |
| 3 | Gemini 3 Pro | 79.7 | 1 | 2026-06-05 | |
| 4 | Claude Sonnet 4.5 | Anthropic | 69.5 | 1 | 2026-06-05 |
| 5 | MiniMax M2.1 | MiniMax | 69.3 | 2 | 2026-08-23 |
| 6 | DeepSeek-V3.2 | DeepSeek | 62 | 1 | 2026-06-05 |
| 7 | MiniMax M2 | MiniMax | 32.8 | 1 | 2026-06-05 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.