Done with openai models, this time for good

I kept coming back to OpenAI switching back and forth between frontier model subscriptions, but I think I am finally done with Openai models. Quick summary is that I am fed up with chatgpt models not following instructions, overengineering things, and introducing bloat that takes many hours to fix rather than getting it right in the first place. I started to feel that I spend more time fixing things that openai models messed up than the efficiency of not having to write boilerplate rust and python code myself. It has absolutely nothing to do with instructions, steering, and guiding. I am a meticulous planner and always write or generate implementation plans and revise and revise until happy. I use top thinking model for that stage and am happy with that portion of work, I actually use different models and competitive products for this process and let certain llms critique the work of other models. That's not the issue. The issues all arise in the implementation portions. Where opus 5.5 (but also previous model versions) or gemini 3.8 follows instructions and plans meticulously, Luna high/xhigh and even sol, whether in 5.6 or now 6 versions add completely unnecessary bloat, wrappers that were never requested, backward bridges and compatibility checks, object versions that drive me insane, and many other artefacts that completely mess up the entire code base. To add to my gripe opus 5.5 is absolutely fantastic, been using it since it came out and am very impressed how thorough it is, the tool usage and agents it spins off, the tests it generates. I don't think I will return to openai ever again, they will introduce newer future and better models, sure. But I think the main issue lies with a bias at openai to rather lie and burry issues in order to appear it solves problems and looks good. I think the philosophy at Anthropic is to be way more thorough, to get it right, even if not being the fastest and needing follow up questions in the planning stage. Openai to me is a company that primarily cares about making a quick buck while anthropic to me is an entitity that still displays the primal passion to output best in class models and to primarily serve the professional community. Hush hush, quick quick vs thorough, organized, adhering to instructions. That's to me what it all comes down to. Add to this the horrific releases of the 6 versions (which very much correlates with everything I said above about rushed releases) and I decided to leave team "fast money"