What is the LLM and setup of choice for code-generation? - Developer Consensus -
I understand that there is clear dominance and liking towards @Claude by AnthropicSonnet 4 for coding. Then there is also open source tools like @Qwen3 Coder performing at near the same levels.
And then there is the integrations itself, do you use them in @Windsurf , @Cursor or maybe even just the CLI tools themselves,
@Gemini CLI offers a great 1000 Free requests and @Claude Code is known to be incredible but for people who can afford it.
And I havent even mentioned the upcoming surplus of new contenders in the space, @Kiro , @Kimi , @Cline , @Github Copilot Agent, @OpenAI Codex?
Is there a somewhat known best practice that the majority of devs are following or are we all just to ourselves? Picking whatever we find best fit, since the difference is intangible?
Replies
I am not sure if other developers use a different approach but personally I use Cursor and let Cursor auto select the LLM model. I have tried the various models but found them to be only marginally different, so I just starting using the auto select mode.