SWE-bench Verified - >4 Hours
Code · 2024-08-13
The >4 hours split of SWE-bench Verified. This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 100.0 |
| Claude Opus 5 | 100.0 |
| GPT-5.6 Sol | 100.0 |
| Opus 4.7 | 66.7 |
| Opus 4.8 | 66.7 |
| Sonnet 5 | 66.7 |
| Composer 2.5 | 66.7 |
| GLM-5.2 | 66.7 |
| GPT-5.5 | 66.7 |
| GPT-5.6 Luna | 66.7 |
| Grok 4.5 | 66.7 |
| Kimi K3 | 66.7 |
| Muse Spark 1.1 | 66.7 |
| Haiku 4.5 | 33.3 |
| Opus 4.6 | 33.3 |
| Sonnet 4.6 | 33.3 |
| DeepSeek-V3.2 | 33.3 |
| Gemini 3.1 Pro Preview | 33.3 |
| Gemini 3.5 Flash | 33.3 |
| Gemini 3.5 Flash-Lite | 33.3 |
| Gemini 3.6 Flash | 33.3 |
| Gemini 3 Flash Preview | 33.3 |
| Gemini 3 Pro Preview | 33.3 |
| GLM-4.7 | 33.3 |
| GLM-5.1 | 33.3 |
| GLM-5 | 33.3 |
| GPT-5.2-Codex | 33.3 |
| GPT-5.2 | 33.3 |
| GPT-5.3-Codex | 33.3 |
| GPT-5.4 Nano | 33.3 |
| GPT-5.6 Terra | 33.3 |
| GPT-5 | 33.3 |
| GPT-5 Mini | 33.3 |
| Grok 4.20 | 33.3 |
| Grok 4.3 | 33.3 |
| Inkling | 33.3 |
| Kimi K2.7 Code | 33.3 |
| Kimi K2 Thinking | 33.3 |
| MiMo-V2.5-Pro | 33.3 |
| MiniMax M2.1 | 33.3 |