Many of the open source coding models are created outside of the US or EU making it challenging to trust when integrating into sensitive or production grade systems.
1) Train a coding model from scratch following SOTA research from Deepseek or Qwen but make the data public and verifiable.
If option 1 is to costly.
2) Post train SOTA Open source model from US or EU (Mistral or OpenAI) using aforementioned coding trainng methods.
Guest
