Token reduction
In short
How many fewer model tokens a page costs as markdown than as HTML. Example: Cloudflare measured 16,180 tokens in HTML against 3,150 in markdown for one post. Related terms: Universal rendering, Subpath llms.txt. Universal rendering: Nuxt term for rendering on the server first and hydrating in the browser.
Source of the example
The example in the summary above is from blog.cloudflare.com, read 4 November 2026.
Where it shows up on this site
Neighbouring terms
- Universal rendering: Nuxt term for rendering on the server first and hydrating in the browser.
- Subpath llms.txt: An llms.txt below the root that covers the pages under its path; v2 says the most specific file applies.
Sources
- blog.cloudflare.com https://blog.cloudflare.com/markdown-for-agents/, read 4 November 2026
Last reviewed 4 November 2026