Unverified50% confidenceFactExact time
自回归生成每个Token都需要完整的前向传播，而输入只需一次编码，这是输出比输入贵的成本结构原因
1
Sources
50%
Confidence
Long-term
Relevance
7/17/2026
First Seen
Sources
Related Claims
Unverified自回归解码具有串行特性，每个Token的生成依赖前一个Token，难以并行78% similarUnverifiedIn coding scenarios, the input side needs to carry a large amount of code context, and the input Token count is often several times that of the output, amplifying costs.75% similarUnverifiedThe computational cost of generating output is typically far higher than that of processing input, and some vendors set separate quotas for input tokens and output tokens.74% similarUnverifiedIn large language model inference, the actual hardware cost of output tokens is far higher than that of input tokens, because the decoding phase is a memory-bandwidth-limited sequential process.68% similarUnverified输出Token定价高于输入Token是因为Prefill阶段可并行矩阵运算而Decode阶段必须串行自回归解码，Decode阶段GPU吞吐量约为Prefill阶段的1/6到1/1068% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/547410API
curl https://kongchang.com/api/v1/knowledge/claims/547410MCP
get_claim(id=547410)