As someone who spends six figures a month benchmarking AI models for hardware engineering, I can tell you: there is no single best model!
Every model has strengths and weaknesses. The best model for one task can be terrible at another.
This is why harnesses like Devin for coding and Flux for hardware are so powerful. They orchestrate multiple models, routing each task to whichever model is best suited for it.
If you’re exclusively using Claude Code, Codex, or any single model, you’re not experiencing the cutting edge of AI.
I think domain-specific harnesses are where the real breakthroughs will happen. Systems that deeply understand the problems they’re solving and know how to get the most out of whatever models and tools are available.
OpenAI has the best model for electronics rn, im using Sol 6.1 all of the time in Backplane