Multi-SWE-bench — Multilingual extension of SWE-bench; base/no-tools variant, large aggregate.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | MiniMax M2.7 | MiniMax | 52.7 | 1 | 2026-08-23 |
| 2 | MiniMax M2.5 | MiniMax | 51.3 | 2 | 2026-08-23 |
| 3 | Claude Opus 4.5 | Anthropic | 50 | 1 | 2026-06-05 |
| 4 | MiniMax M2.1 | MiniMax | 49.4 | 2 | 2026-08-23 |
| 5 | Claude Sonnet 4.5 | Anthropic | 44.3 | 3 | 2026-06-06 |
| 6 | Gemini 3 Pro | 42.7 | 1 | 2026-06-05 | |
| 7 | Kimi K2 (Reasoning) | Moonshot | 41.9 | 4 | 2026-06-06 |
| 8 | GPT-5 | OpenAI | 39.3 | 1 | 2026-05-15 |
| 9 | DeepSeek-V3.2 | DeepSeek | 37.4 | 3 | 2026-06-06 |
| 10 | MiniMax M2 | MiniMax | 36.2 | 3 | 2026-08-23 |
| 11 | Claude Sonnet 4 | Anthropic | 35.7 | 2 | 2026-06-06 |
| 12 | Qwen3 Coder 480B A35B Instruct | Alibaba | 32.7 | 2 | 2026-08-23 |
| 13 | GLM-4.5 | Z.ai | 31.7 | 1 | 2026-05-15 |
| 14 | K2-Instruct-0711 | Moonshot | 31.3 | 1 | 2026-05-15 |
| 15 | GLM-4.6 | Z.ai | 30 | 1 | 2026-06-06 |
| 16 | DeepSeek-V3.1 | DeepSeek | 29 | 1 | 2026-05-15 |
| 17 | Seed Oss 36B Instruct | ByteDance | 17 | 1 | 2026-06-15 |
| 18 | Qwen3 30B A3B | Alibaba | 9.5 | 1 | 2026-06-15 |
| 19 | Qwen3 32B | Alibaba | 7.7 | 1 | 2026-06-15 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.