I assume this is a typo on the Model pricing page, the cached price should be 7 requests.
It’s a massive difference in price, I can’t see any way it’s not a mistake.
It was changed to 9.3. What I don’t get though is
- Is a continuation after a tool call another request which includes all my previous context, of the same length as the first request?
- The first table here says most of it was cached, but still says it’s counted as “2.7 requests” - which means even though it is all cached, I still pay the 2.7 requests? Why? It should be the cached tokens count * cached price, is it not?
- There is a “deepseek-v3.1” that is not deletable and “deepseek-v3” which is deletable in my cursor. But in docs, I see deepsek v3 is named “deepseek-v3” and no “deepseek-v3”. Why are some models deletable and some not?
- What does “112.5 requests / MTok per hour” mean?
