> I'm starting to wonder if this increased token usage is inadvertently bleeding into how Anthropic actually trains their model
Related: Sonnet 5’s new tokenizer increases token usage by 30%. (https://simonwillison.net/2026/Jun/30/claude-sonnet-5/)
> I'm starting to wonder if this increased token usage is inadvertently bleeding into how Anthropic actually trains their model
Related: Sonnet 5’s new tokenizer increases token usage by 30%. (https://simonwillison.net/2026/Jun/30/claude-sonnet-5/)