Unverified50% confidenceFactExact time
When 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.
1
Sources
50%
Confidence
Long-term
Relevance
8/17/2026
First Seen
Sources
Inferock Bench：为每次LLM调用开一张独立账单的开源成本审计工具
Related Claims
UnverifiedEach retry resends the complete prompt input tokens and is billed in full; a request with 4,000 input tokens retried 3 times will consume 16,000 tokens.68% similarUnverifiedOutput tokens (content generated by the model) are typically more expensive than input tokens (prompts provided by the user).65% similarUnverifiedThe API cache hit mechanism allows input tokens that have already been sent to enjoy discounted pricing on repeated calls, typically at 10%–50% of the original price.64% similarUnverified当API返回速率限制、超时或服务端错误时客户端自动重试，提供商可能已对已生成token计费导致重试产生额外成本63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/764521API
curl https://kongchang.com/api/v1/knowledge/claims/764521MCP
get_claim(id=764521)