Kimi · Pricing pageSource update

Summary

The model inference interface uses usage-based pricing for both Input and Output. For the K3 series models, cache writes are billed separately at TTL (5min / 1h); inputs that are cache hits are billed only at the cache hit price, without a separate cache write charge. If you upload and extract document content and pass the extracted content as Input to the model, the document content will also be billed on a per-use basis. File-related APIs (document content extraction / file storage) are free for a limited time, meaning the file APIs themselves do not incur fees when you only upload and extract documents.