Unverified50% confidenceFactExact time
DeepSeek、Qwen、GLM三家都在旗舰之外单独推出高速轻量的Flash版，且Flash版下载量普遍更高
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/10/2026
First Seen
Valid until: 12/9/2026
Sources
Related Entities
Related Claims
UnverifiedFlash版轻量模型的下载量普遍高于旗舰版，反映用户更看重能在本地显卡运行的模型66% similarPartially VerifiedQwen3 Next Flash的4-bit量化版本（Q4K XL）约111GB65% similarUnverifiedDeepSeek V4 Flash的UD-Q8_K_XL量化版本文件体积为162GB，采用MXFP4+BF16混合精度，达到完全无损65% similarUnverifiedQwen3.8-Flash-Next的1-bit Dynamic GGUF版本可在75GB内存或统一内存上运行，相比原始BF16格式体积缩小79%，保留80%的top-1准确率64% similarUnverifiedQwen3.8 Flash Next 原生上下文窗口为26.2万token，借助YaRN技术可扩展至100万token62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/888672API
curl https://kongchang.com/api/v1/knowledge/claims/888672MCP
get_claim(id=888672)