Unverified85% confidenceFactTime unknown
DeepSeek V4 Flash的KV Cache显存降至V3.2版本的7%(降低13.7倍)
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
6/2/2026
First Seen
Valid until: 8/31/2026
Sources
DeepSeek V4技术深度拆解:百万Token与极致性价比
bilibili码书z
Related Entities
Related Claims
Unverified经过压缩后,DeepSeek V4 Flash模型体积从接近500GB降至80.8GB77% 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74% similarVerifiedDeepSeek V4 Flash拥有284B总参数,每次激活13B,显存需求约150G73% similarUnverifiedDeepSeek-V4 Pro版参数达1.6T,每次推理激活49B;Flash版参数284B,激活13B72% similarUnverifiedDeepSeek V4 Flash的API调用成本通常只有满血版的几分之一72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/36598API
curl https://kongchang.com/api/v1/knowledge/claims/36598MCP
get_claim(id=36598)