Unverified50% confidenceBenchmarkExact time
在MI355X上通过移除不必要的专家填充，DeepSeek-V4的FP4 MoE模型权重从159.07 GB降至112.36 GB
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
Unverified经过压缩后，DeepSeek V4 Flash模型体积从接近500GB降至80.8GB76% similarUnverified经过SVDQuant的4-bit量化后，FLUX模型显存需求可从约24GB降至约6GB75% similarUnverifiedllama.cpp项目使27B参数模型的显存需求从FP16约54GB降至Q4量化约16GB74% similarUnverified140GB的FP16 70B模型经4-bit量化后可压缩至约35-40GB，通信开销降低约75%；2-bit量化可压缩至约18GB但带来精度损失74% similarUnverifiedM3原始权重存储下限：BF16约854GB，8-bit约428GB，4-bit约214GB72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/897571API
curl https://kongchang.com/api/v1/knowledge/claims/897571MCP
get_claim(id=897571)