https://www.claudecodecamp.com/p/i-measured-claude-4-7-s-new-tokenizer-here-s-what-it-costs-you
Article
-
Analysis of Opus 4.7’s new tokenizer showing ~30% more tokens for equivalent code/prose.
-
Higher input token counts translate directly to higher API costs per request.
-
GitHub Copilot’s internal multiplier for 4.7 jumped from 3x to 7.5x.
-
Author argues cost-per-task (not token count) is the correct metric but data is thin.
Discussion
-
namnnumbr and tabbott: token count is wrong metric—cost-per-task and human review time matter more.
-
yuanzhi1203: discovered Anthropic A/B tested routing 4.6 calls to 4.7 silently two weeks prior.
-
montjoy: comparison should be 4.7 xhigh vs 4.6 max, not max vs max—Anthropic recommends xhigh now.
-
Multiple users reporting hitting weekly/5-hour limits dramatically faster than with 4.6.
Discuss on HN
