Unverified50% confidenceFactExact time

千问3.8 27B是27.8B的稠密模型，支持图像和视频原生多模态输入，原生262K上下文

1

Sources

50%

Confidence

Long-term

Relevance

8/18/2026

First Seen

Sources

千问3.8 27B实测：一张显卡跑长程编程Agent

Related Claims

Unverified一张100KB的图片经Base64编码后体积会增加约33%，变为约133KB的文本数据72% similar Verified32B参数规模的模型在量化后通常可以压缩到约16-20GB显存占用，一张NVIDIA RTX 4090（24GB显存）就能运行71% similar Unverified千问3.6支持256K超长上下文70% similar Unverified256GB起步版本对于4800万像素ProRAW拍摄和4K ProRes视频录制而言空间紧张，重度影像用户建议直接选512GB或更高70% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/768401

API

curl https://kongchang.com/api/v1/knowledge/claims/768401

MCP

get_claim(id=768401)