Qwen2.5-Coder-32B-Instruct

Qwen2.5-Coder-32B-Instruct is the chat-oriented coding variant in the Qwen2.5-Coder line. It was developed for code completion, debugging, explanations, and code generation. Instruct fine-tuning makes the checkpoint usable directly through a chat interface. The underlying Coder series was trained on 5.5 trillion code tokens and covers many programming languages.

With 32 billion parameters and 131,072 context tokens, the checkpoint handles larger files and repository excerpts than Qwen2.5-Coder 0.5B. The current models.dev catalog lists 8,192 output tokens for this serving profile. That is a host configuration, not a universal limit of the open weights.