Xiaomi Mimi v2.5 Pro token plan Cache charges
Hello, you appear to be charging way too much for cache hits on your token plans. In fact I would say it’s really bad.
It’s so bad that I actually went through 2 months of tokens with very little effort, just simple agentic work.
Your press release and blogs talk about how the models are designed for agentic workflow but it doesn’t look like your billing of cache is.
In fact when I did the math it was much cheaper for me to use the API than the token plan! Why is that possible?
Why do you charge the same for cache hits as you do for output on the token plan!?
Also why is this not clearly documented anywhere so it’s obvious to everyone that choosing the token plan for agentic workflow and what Xiaomi says the model was built for would be prohibitively expensive?