Unverified50% confidenceConceptExact time
Every piece of text in a context file is injected into the model's context window on every call, causing increased token cost, i.e., the "context tax".
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAPI costs stem from a per-token billing mechanism, and large code generation tasks consume a large number of output tokens, causing costs to accumulate.72% similarUnverifiedCurrent mainstream model pricing is split into input Tokens and output Tokens, with output Tokens typically costing 2-4x more than input Tokens69% similarUnverifiedA token is the basic unit by which LLMs measure text, roughly corresponding to half to one English word, and the computational cost of a model is billed by the number of tokens.69% similarUnverifiedThe cost of each model call consists of two parts: input tokens (Prompt) and output tokens (Completion).67% similarUnverified每百万Token的价格并非可比单位，因为生成同一段文本不同模型消耗的Token数差异巨大，真正该关注的是每次成功结果的总成本67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1036974API
curl https://kongchang.com/api/v1/knowledge/claims/1036974MCP
get_claim(id=1036974)