Mixtral is a french company creating an open source coding model that looks promising.
https://mistral.ai/news/codestral/
Personally I trust them and LLaMa more than DeepSeek or Qwen since I’ve heard a rumor that chinese models may intentionally guide users to add vulnurabilities into their software to make them more exploitable by the PLA. And that they also repeat CCP’s disinformation narratives, which can be dangerous to national security if left unchecked.