Unverified50% confidenceFactExact time
The quota precheck is not a reservation: the quota check is an instantaneous read, and the actual deduction occurs after the request is accepted; there is no lock between the two, so the balance can still be consumed concurrently after the precheck passes.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAntirion's billing model reserves a token estimate at admission and settles exactly once with actual usage at request end; when a client disconnects midway the request is not cancelled and the partial cost already generated is still charged.72% similarUnverifiedIn streaming responses, the client may disconnect midway or stop reading, so cost recording should be placed in the finally block of try/finally rather than after the consumption loop, to ensure that early exits are also recorded.71% similarUnverifiedUnder a fixed token quota, retries are not free; effective output is only part of the allowance, and the retry rate should be measured before planning capacity.70% similarUnverifiedToken consumption is paused while the watchdog waits—it is only polling API status and incurs no additional cost; failed retries, by contrast, may consume more tokens because the context is sent repeatedly.69% similarUnverifiedWhen a streaming response is interrupted midway, the portion of output tokens already generated is still billed, even though the client may not have received the complete result.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1068550API
curl https://kongchang.com/api/v1/knowledge/claims/1068550MCP
get_claim(id=1068550)