[KongchangAI]
Unverified50% confidenceFactExact time

截至2025年，主流frontier模型（如GPT-4o、Claude 3.5 Sonnet、Gemini 1.5 Pro）的API推理成本约在每百万输入token 2-15美元区间，输出token成本通常为输入的3-5倍

1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/13/2026
First Seen
Valid until: 10/11/2026

Sources

Perplexity Max额度从40K缩至10K，付费用户该怎么办？
redditr/perplexity_ai7/12/2026

Related Claims

Unverified主流模型推理成本以每百万token计价：GPT-4o约为5-15美元，Claude 3.5 Sonnet约为3-15美元，多步骤代理任务的token总量往往是单次对话的10-50倍85% similarVerified以GPT-4级别的模型为例，每百万输入token的费用可达数美元77% similar

Cite This Claim

Stable URI
https://kongchang.com/claim/502186
API
curl https://kongchang.com/api/v1/knowledge/claims/502186
MCP
get_claim(id=502186)