I don’t think it’s API costs. Their Sonnet 4.6 is just 1x premium request which matches the 1x cost of the various GPT Codex models.

Sonnet is the worse model though, therefore it's expected that it is cheaper, the comparison would be Opus and GPT. That Anthropic's worse model is the same request cost as the best OpenAI model is what I mean when talking about Microsoft flexing their partnership.