AI BriefingOpus 4.7 Token Usage Increase
Key point
In the same email classification agent, Opus 4.7's token usage and cost increased.
Details
After running Opus 4.6 for 3 days, the same email classification agent was switched to Opus 4.7 for 3 days for comparison.
The conditions kept SOUL.md, skills, and a daily email volume of about 180 items identical.
- input tokens: 847k/day → 1,091k/day (+29%)
- output tokens: 123k/day → 189k/day (+54%)
- daily cost: $7.31 → $10.18 (+39%), an increase of about $86 per month
The input increase was explained by a tokenizer change, and the output increase by an increase in reasoning tokens created by adaptive thinking in later turns.
Since there was almost no quality difference for the email classification use case, they reverted to 4.6.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Analysis of the 'Inference Gap' Between Claude's Reduced Reasoning Tokens and Benchmarks
US Appeals Court Upholds Pentagon Blacklisting of Anthropic as Supply Chain Risk