01
Official API or machine-readable source
Structured provider-owned pricing sources are preferred when available.
The tracker favors official, narrow, and low-frequency sources. HTML parsing is treated as a cautious fallback, not a default contract.
01
Structured provider-owned pricing sources are preferred when available.
02
Provider documentation and official pricing pages are used when structured sources are not available.
03
HTML parsing is a last resort and is kept narrow, low-frequency, and conservative.
No broad crawling.
No login-required content.
No challenge, CAPTCHA, authentication, or rate-limit bypass.
No claim that robots.txt alone makes a source appropriate.
OpenAI and Anthropic are currently treated as caution-level official pricing page sources. That means the pricing pages are useful official references, but they are not treated as stable machine-readable interfaces.
Provider-specific pricing concepts may require normalization. For example, Anthropic prompt caching read pricing is displayed with a cached input caveat where relevant.