Does reasoning effort work with a local DeepSeek model behind llama-server?

This answer refers to the version in the original report. Check the controls and behavior in your own version.

Compare the adapter's outgoing fields with your llama-server version and model template. Send two short requests with the same model and prompt, varying only the effort setting. Inspect redacted requests and server logs before concluding the setting is supported.

References

• Original discussion · GitHub Discussions