Unverified60% confidenceFactExact time
In FP16 format each parameter occupies 2 bytes, while in INT4 format it occupies only 0.5 bytes, theoretically achieving a 4x compression ratio.
2
Sources
60%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
Alibaba Bans Claude Code: The Battle Over AI Coding Tool Security and Data Sovereignty
bilibiliSynthMindX7/4/2026
Related Claims
UnverifiedTaking the Q4_0 format as an example, every 32 FP32 weights are packed into one quantization block containing one scale factor and 32 4-bit integers, averaging approximately 4.5 bits per parameter, reducing memory usage to approximately 1/8 of the original.80% similarPartially VerifiedNVFP4 occupies 4 bits per parameter, i.e., 0.5 bytes, theoretically halving weight memory usage and load bandwidth compared with FP8.78% similarUnverifiedFP16使用16位存储数值（1位符号位、5位指数位、10位尾数位），可表示约65000个离散值75% similarUnverifiedGGUF格式支持将FP32或FP16浮点权重压缩至2位到8位整数74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/489176API
curl https://kongchang.com/api/v1/knowledge/claims/489176MCP
get_claim(id=489176)