Description
Thinking models like DeepSeek-R1 emit thinking tags in the output. Is there a way to filter these out easily? Currently they make it directly into the output and so mess up metrics.
{{ message }}
No one assigned
No projects
No milestone
None yet
No branches or pull requests