Unverified50% confidenceFactExact time
Meta provides two checkpoint formats: full-precision BF16 (about 1.6TB) and a compressed NVFP4 version (about 530GB).
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/16/2026
First Seen
Valid until: 1/14/2027
Sources
Related Entities
Related Claims
UnverifiedMeta Llama 3的405B参数版本以BF16精度加载需约810GB内存，量化至INT4精度需约200GB77% similarUnverifiedMeta simultaneously offers two 4-bit Dynamic quantization versions: the larger version occupies 19.7 GB of VRAM (precision loss of approximately 0.2%), and the smaller version occupies 16.8 GB of VRAM (precision loss of approximately 1%).73% similarUnverifiedThe UD-Q8_K_XL quantized version of DeepSeek V4 Flash has a file size of 162 GB, uses MXFP4+BF16 mixed precision, and achieves completely lossless quality.67% similarVerifiedA GPT-4-class model with 175 billion parameters requires approximately 175 GB of VRAM even when stored at fp8 precision.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1042238API
curl https://kongchang.com/api/v1/knowledge/claims/1042238MCP
get_claim(id=1042238)