Interesting note from Simon Willison that Claude’s new tokenizer translates to higher costs:

Opus 4.7 uses the same pricing is Opus 4.6 - $5 per million input tokens and $25 per million output tokens - but this token inflation means we can expect it to be around 40% more expensive

Anthropic pricing is already much more expensive than OpenAI. With how good the new Codex app for Mac is, I think we’ll see more developers move to Codex. But people are also comfortable with their workflows, so there’s not going to be a quick shift, and models and features change nearly every month.