I think this model looks interesting. Jetbrains has lots of experience with software development. I think it worth a look:
Its a 12b model, which they are comparing it to other models in size class. Does anyone here code with Qwen 3.5 9b or Mistral 14b, might be useful for RL – is that what you are thinking?
I am coding some AI agents for ISO work products review. I plan to run it in comparison to Qwen 3.5 122B and Qwen 3.6 to see how they perform during agent development and running agents during ISO work products creation.
I am trying to see which models might be good candidates to integrate domain knowledge.
I mainly use Google ADK and LangGraph.