Leading LLMs from around the world, plus image, video and audio generators — all called through one OpenAI-compatible API. Input tokens that hit the upstream cache are billed at the cache rate (as low as 30% of list). Prices follow each vendor's official public rates and are being filled in progressively.