2026
cobol-coder
A Qwen2 0.5B fine-tune for COBOL system understanding, trained with Unsloth and Hugging Face Transformers.
A small model that reads COBOL
Qwen2 0.5B
Base model
BF16
Precision
Detail
- Part of a broader effort to give language models useful competence on legacy mainframe code, where training data is scarce and the cost of a wrong answer is high.
- Companion to a rebuilt evaluation harness — the fine-tune and the thing that measures it were built together, deliberately.
- Sits alongside larger fine-tunes of Llama 3.2, Granite, and DeepSeek for the same domain.
Stack
Underlined entries link to the writing on that topic.