Unverified50% confidenceFactExact time
Under FP4/INT4 quantization, the model size is approximately one-quarter of that under FP16.
1
Sources
50%
Confidence
Long-term
Relevance
9/29/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedINT4量化将原本由65536个不同值（FP16）表示的权重压缩到仅16个离散级别79% similarVerifiedFP32每个参数占4字节，FP16/BF16占2字节，INT8量化占1字节，INT4量化占0.5字节78% similarUnverifiedW4A4量化将权重和激活值均压缩至4位整数，理论上相比FP16可将显存占用降低至约1/477% similarUnverifiedThe performance gap between a carefully quantized 4-bit model and the original FP16 model is typically within 1–3 percentage points.75% similarUnverifiedThe FP4 format has only 4 bits (1 sign bit, 2 exponent bits, 1 mantissa bit) and can precisely represent only approximately 16 discrete values.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/957937API
curl https://kongchang.com/api/v1/knowledge/claims/957937MCP
get_claim(id=957937)