Unlike Apple, OpenAI doesn't have nearly the same moat. The Chinese labs are going to eat their lunch at this rate.

They do have the psychological cachet of Apple though – if Apple is the reasonably polished, general-purpose consumer device company to the average punter, OpenAI has a reputation of being the "consumer AI" company to the average punter that's hard to dislodge.

Anthropic has shut them off from API access, so the most interesting comparison wouldn't be there anyways.

Presumably because GLM 4.5 or Qwen3 comparisons would clobber them in eval scores.

You can check the same evals OpenAI used for those models

Hint: unclobbered