Unverified50% confidenceBenchmarkExact time
On MI355X, by removing unnecessary expert padding, the FP4 MoE model weights of DeepSeek-V4 were reduced from 159.07 GB to 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
UnverifiedAfter compression, the DeepSeek V4 Flash model size was reduced from nearly 500 GB to 80.8 GB.76% 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% similarUnverifiedA 140 GB FP16 70B model can be compressed to approximately 35–40 GB after 4-bit quantization, reducing communication overhead by about 75%; 2-bit quantization can compress it to approximately 18 GB but introduces accuracy loss.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)