Code Migration - COBOL
Code · 2026-06-16
The COBOL split of Code Migration. 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 |
|---|---|
| Opus 4.7 | 70.0 |
| Claude Opus 5 | 70.0 |
| GLM-5.2 | 70.0 |
| GPT-5.5 | 70.0 |
| GPT-5.6 Luna | 70.0 |
| GPT-5.6 Sol | 70.0 |
| Grok 4.5 | 70.0 |
| Opus 4.8 | 68.6 |
| Sonnet 4.6 | 68.6 |
| Sonnet 5 | 68.6 |
| GPT-5.4 | 68.6 |
| Muse Spark 1.1 | 68.6 |
| GPT-5.6 Terra | 65.8 |
| Gemini 3.6 Flash | 65.7 |
| Kimi K2.7 Code | 61.3 |
| DeepSeek-V4-Pro | 58.4 |
| Kimi K2.6 | 57.1 |
| Gemini 3.5 Flash | 55.5 |
| GLM-5.1 | 46.6 |
| MiMo-V2.5-Pro | 44.3 |
| MiniMax M3 | 43.9 |
| Claude Fable 5 | 40.0 |
| Gemini 3.1 Pro Preview | 36.8 |
| MiMo-V2.5 | 27.3 |
| GPT-5.4 Mini | 27.1 |
| Inkling | 26.1 |
| Gemini 3.5 Flash-Lite | 7.0 |
| MiniMax M2.7 | 4.3 |
| Gemini 3.1 Flash-Lite Preview | 1.3 |
| Qwen3.6 Plus (2026-04-02) | 1.3 |
| Haiku 4.5 | 0.0 |
| Gemini 3 Flash Preview | 0.0 |
| Grok 4.3 | 0.0 |