I wanted a way to track the token economy the way you'd follow equities or other financial markets: who's gaining market share, who's cutting prices, which new models are actually getting used. OpenRouter publishes a lot of the live data, but it's spread across various pages and endpoints, mostly as snapshots, so I started collecting and organizing it.
The historical data was the hard part. OpenRouter doesn't publish an archive, so I rebuilt it from the Wayback Machine. I indexed about 2 million archived OpenRouter URLs, downloaded close to a million captures, and wrote parsers for every page and API format the site has used since 2023. Most of the work was in reconciling the disparate feeds of data, iteratively building up and infilling the dataset via derivation.
The historical data was the hard part. OpenRouter doesn't publish an archive, so I rebuilt it from the Wayback Machine. I indexed about 2 million archived OpenRouter URLs, downloaded close to a million captures, and wrote parsers for every page and API format the site has used since 2023. Most of the work was in reconciling the disparate feeds of data, iteratively building up and infilling the dataset via derivation.
Hopefully some people might find it useful.