Cloudflare cuts AI code-review cost 68% to $0.77
Original title: Cutting our inference bill by 68% by being obsessive about optimization
Median review time fell from 4 minutes 9 seconds to 1 minute 32 seconds after the changes.
Cloudflare said its internal AI code reviewer rose from about $1.19 in April to about $2.46 by the end of August, largely after moving reviewers onto Anthropic models. Work starting in mid-September cut the average high-effort review to $0.77, 68% below $2.43 in the two costliest weeks. About 40% of the savings came from reviewer code changes and the rest from model and effort experiments. The reviewer now runs on OpenCode v2.
- Switching to Anthropic models raised cost per job from $1.38 to $2.03.
- Codex on Claude Opus 4.7 rose from about $0.25 to $0.88 per review.
- Tokens per review grew from 0.91 million to 2.99 million.
- The cache hit rate rose from 85.7% in April to 97% by September.
- v1 and v2 agreed on approval in 98% of tested merge requests.
- Vendors
- Anthropic
- Products
- OpenCode
- Organizations
- Cloudflare
- AI models
- Claude Opus 4.7
From the original article
14 min read at blog.cloudflare.com
When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intelligent model routing to slash our inference bill by 68% while speeding up reviews.
By Ryan Skidmore · blog.cloudflare.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.