Unverified50% confidenceFactExact time

Anthropic的新模型代号为Fable，其运行成本高于Opus模型

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

7/13/2026

First Seen

Valid until: 10/11/2026

Sources

月付200美元用不上旗舰模型？Claude订阅困局与Token计费方案探讨

redditr/ClaudeAI7/12/2026

Related Claims

UnverifiedAnthropic发布了Fable 5模型，该模型在工作时消耗两倍于Opus 4.8的Token数量66% similar Unverified单价更高的大模型在完整任务上的总成本反而可能更低65% similar Unverified两个智能得分相近的模型其推理成本可能相差数倍甚至一个数量级64% similar Unverified对万亿参数模型而言，处理长上下文时KV Cache的显存开销甚至可能超过模型权重本身64% similar Unverified推理增强型模型如o1系列的价格可达GPT-4o的数倍甚至十倍以上62% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/502123

API

curl https://kongchang.com/api/v1/knowledge/claims/502123

MCP

get_claim(id=502123)