Unverified50% confidenceFactExact time
The DeepGEMM-Ascend library supports BF16 (16-bit), FP8 (8-bit), and FP4 (4-bit) numeric formats.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/3/2026
First Seen
Valid until: 1/1/2027
Sources
Related Entities
Related Claims
UnverifiedOn the Huawei 950DT chip, DeepGEMM Ascend edition achieves a hardware utilization rate of 99.8% for BF16 multiplied by BF16.70% similarUnverifiedThe BF16 weight files for Naive-N0.5-Flash are approximately 618GB, and the FP8 format version weights are approximately 315GB67% similarVerifiedFP8 is an 8-bit floating-point format natively supported by newer architectures such as the NVIDIA H100, which can compress data volume by half compared to FP16/BF1666% similarUnverifiedGGUF格式支持将FP32或FP16浮点权重压缩至2位到8位整数66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/970595API
curl https://kongchang.com/api/v1/knowledge/claims/970595MCP
get_claim(id=970595)