Opencode and the MiniMax models are good enough

I’ve dabbled with the MiniMax models (and Kimi/GLM) over the last few days.

They are good enough for 90% of tasks. Codex and Claude still have them beat but I think you only nneed those for planning now. Execution could be MiniMax (or GLM 5.2) and you’ll find them pretty good.

Caveat here is if you one-shot everything you are going to get burned but thats on you for being a lazy arsehole not the model.

Opencode with my claude settings so that it uses all the various skills I have setup is also very handy. If I’m honest I think I prefer Claude Code but it might be a case of familiarity over performance.

Tags:

#ai