Unverified50% confidenceFactExact time
During Transformer model inference, input tokens can be processed in parallel (prefill stage), while output tokens must be generated one by one autoregressively (decode stage), causing the unit price of output tokens to typically be higher than that of input tokens.
1
Sources
50%
Confidence
Long-term
Relevance
9/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTransformer 架构下模型在推理时仅处理当前输入的 token 序列，不存在跨请求的内部状态持久化76% similarVerifiedThe token generation process of large language models is serial, based on an autoregressive mechanism, where each token's generation depends on the previous token.74% similarUnverifiedEach model invocation calculates both input tokens and output tokens simultaneously, and both together determine the cost of a single call.73% similarUnverified在Transformer自回归推理模式下，输出Token数量与计算开销之间存在超线性关系73% similarUnverifiedThe autoregressive generation characteristic of LLMs determines that their output is streamed token by token, stemming from the decoding strategy of the Transformer architecture.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/855688API
curl https://kongchang.com/api/v1/knowledge/claims/855688MCP
get_claim(id=855688)