Unverified50% confidenceFactExact time
For coding agents, input tokens far outnumber output tokens because the model must repeatedly read codebase context, conversation history, tool call return results, and error logs.
1
Sources
50%
Confidence
Long-term
Relevance
9/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn programming assistant scenarios, system prompts and code context typically consume a large number of tokens, resulting in an extremely high cache hit rate.72% similarUnverifiedMainstream large model APIs charge separately for input tokens and output tokens, with input tokens typically including the entire conversation history (i.e., the context).71% similarUnverifiedIn the Prompt Cache mechanism, when the prefix content of an API request is identical to a previous one, the system reuses the already-computed KV cache to complete cache reads at a lower price, while cache write is the billing item for the first time content is written into the cache.66% similarUnverifiedIn agent scenarios, due to multiple rounds of tool calls, each call must carry the previous conversation history and system prompt, causing input tokens to grow cumulatively, and a code refactoring task may consume tens of thousands or even hundreds of thousands of tokens.66% similarUnverifiedAPI billing covers both input tokens and output tokens, and as a conversation extends, the input tokens carried in each request grow linearly.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/926829API
curl https://kongchang.com/api/v1/knowledge/claims/926829MCP
get_claim(id=926829)