In the team pricing of Cursor, there are some doubts regarding the consumption of requests: The request consumption in the max mode and the agent_review mode is not an integer. Is this calculated based on token consumption and cache hits? Additionally, I noticed in the official documentation that there seems to be no clear explanation regarding the request consumption of the model in the team mode. For instance, the requests for sonnet4.5 and opus4.5 are both 2, while the request for gpt-5.2 is 1, etc.
Hey there,
For users on a request-based plan, the fractional values appear when Max Mode is enabled. Max Mode switches from fixed per-request pricing to token-based billing under the hood. The actual token usage (including any cache hits that reduce your costs) gets converted back to “requests” for display. The agent_review mode also uses token-based pricing, which is why you see fractional values there, too.
For model costs, some premium models count as more than 1 request. You can find more details about this in the docs on Request-Based Legacy Pricing!
1 Like
Thank you for your explanation. I had seen that document before, but because some of the requested consumption did not match the actual consumption, I previously thought that the document was out of date.