Anyone using Open Source LLMs for Code Generation?
Do any of you work with open source models for code generation?
I actually only use Claude 4.5 Sonnet as a coding agent. It seems to work best for me by far, but it's not really "indie" if you make yourself dependent on a proprietary provider. I would love to work with a free LLM. Does anybody have any experience with this?
Thank you very much for sharing!
Clear and practical, thanks. Did anything surprise you along the way?
I've been testing multiple open source models while building PromptProbe , a tool for measuring prompt reliability across repeated runs. My experience is that many open source models are surprisingly capable for code generation , but consistency is where they still lag behind claude. A model can solve the task once , but repeating the same prompt several times often produces noticeably different outputs. Curious if anyone here has found an open source that's both strong and consistently reliable for coding.