About Toolsummary
A daily record of what LLM APIs cost, kept in the open because a price is the one fact a language model cannot reconstruct from memory.
What this is
Toolsummary tracks published per-token pricing for LLM and inference APIs and records a snapshot every day. That log is the product: it turns "what does this cost" into a fact with a date on it, rather than a number someone typed over.
How prices are collected
Prices come from provider APIs and public price lists, normalised to USD per million tokens. Anything ambiguous goes to a human review queue instead of onto a page. A rejected row never enters the history.
How fresh is this?
Every page carries the date its data was pulled. The pipeline runs daily at 04:30 UTC; a page showing an older date than today means that day's refresh has not landed yet.
Independence
No provider pays for placement, and no row is priced differently from any other. If a number is wrong, the fastest fix is the submit form.
Using the data
The current dataset is available as JSON at /api/prices.json. The blog has an RSS feed.