SWE-bench Verified (Vals 1–4 hour split)
Code · 2026-06-17
Vals difficulty split containing SWE-bench Verified tasks estimated to take 1 to 4 hours for humans. It preserves the duration split rather than aggregating all rows under the parent benchmark.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 97.6 |
| Claude Fable 5 | 92.9 |
| Kimi K3 | 92.9 |
| Claude Opus 5 | 90.5 |
| GPT-5.6 Luna | 85.7 |
| Sonnet 5 | 76.2 |
| Opus 4.8 | 73.8 |
| Grok 4.5 | 73.8 |
| GLM-5.2 | 66.7 |
| Opus 4.7 | 64.3 |
| GPT-5.6 Terra | 64.3 |
| Inkling | 57.1 |
| GPT-5.3-Codex | 54.8 |
| Composer 2.5 | 52.4 |
| Gemini 3.5 Flash | 52.4 |
| Muse Spark 1.1 | 52.4 |
| Sonnet 4.6 | 50.0 |
| Gemini 3.6 Flash | 50.0 |
| GPT-5.4 | 50.0 |
| GPT-5.5 | 50.0 |
| Kimi K2.7 Code | 50.0 |
| MiniMax M3 | 47.6 |
| DeepSeek-V4-Pro | 45.2 |
| GLM-5.1 | 45.2 |
| Opus 4.6 | 42.9 |
| Gemini 3.1 Pro Preview | 42.9 |
| Gemini 3 Pro Preview | 42.9 |
| Opus 4.5 | 40.5 |
| Gemini 3.5 Flash-Lite | 40.5 |
| Kimi K2.6 | 40.5 |
| MiniMax M2.1 | 40.5 |
| Muse Spark | 40.5 |
| Gemini 3 Flash Preview | 38.1 |
| GPT-5.2 | 38.1 |
| Grok 4.3 | 38.1 |
| MiniMax M2.5 | 38.1 |
| Nemotron 3 Ultra 550B A55B | 38.1 |
| Qwen3.6 Plus (2026-04-02) | 38.1 |
| Qwen3.7-Max | 38.1 |
| GLM-4.7 | 35.7 |