[KongchangAI]
Unverified50% confidenceTradeoffExact time

端侧模型推理需为KV缓存、中间激活值和系统开销预留内存，实际可支撑的上下文长度可能远低于云端版本的128K窗口

1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen

Sources

27B Parameter LLM on an iPhone? A Deep Dive into On-Device Full-Parameter AI

Related Claims

Unverified长上下文在首次推理时需完整计算KV Cache，消耗GPU计算资源并占用显存带宽，拉长首token延迟（TTFT）70% similarUnverifiedFor a model with 70B parameters, the KV Cache for an 8K context may occupy several additional gigabytes of VRAM69% similar

Cite This Claim

Stable URI
https://kongchang.com/claim/584673
API
curl https://kongchang.com/api/v1/knowledge/claims/584673
MCP
get_claim(id=584673)