SWE-Dev — Distinct named SWE-Dev benchmark, Kimi K2 Instruct card, own scale (53-67)
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4 | Anthropic | 67.1 | 1 | 2026-05-15 |
| 2 | Kimi K2 (Reasoning) | Moonshot | 66.6 | 1 | 2026-05-15 |
| 3 | Qwen3 Coder 480B A35B Instruct | Alibaba | 64.7 | 1 | 2026-05-15 |
| 4 | GLM-4.5 | Z.ai | 63.2 | 1 | 2026-05-15 |
| 5 | K2-Instruct-0711 | Moonshot | 61.9 | 1 | 2026-05-15 |
| 6 | DeepSeek-V3.1 | DeepSeek | 53.3 | 1 | 2026-05-15 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.