You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The LLM you are using "decides" what to do, not openhands. I don't really know how good for agents is the specific DeepSeek Lite GGUF you're using.
There's a very significant difference between LLMs and how well they work with the code and code history, for an agent to succeed tasks. We currently recommend Sonnet 3.5 (new) for best results.
Please take a look at this discussion for information on LLMs we've evaluated. You may see that DeepSeek is in the best case we found at 7%, compared to Sonnet at over 40% success.
This will change, but as of now it is very relevant for your tests, I'm afraid.
Is there an existing issue for the same bug?
Describe the bug and reproduction steps
https://www.all-hands.dev/share?share_id=2ccf1523f77b5bdf9ce6db8f3f9a8e7c62f73b0b21453744fc20882a0c5445b9
LLM: https://huggingface.co/lmstudio-community/DeepSeek-Coder-V2-Lite-Instruct-GGUF
OpenHands Installation
Docker command in README
OpenHands Version
0.15.3
Operating System
WSL on Windows
Logs, Errors, Screenshots, and Additional Context
No response
The text was updated successfully, but these errors were encountered: