Compact models·EASYHUB JOURNAL
BitCPM-CANN opens low-bit models trained on Ascend
What changed
The release spans compact ternary models and a low-bit training approach on Ascend. Native low-bit training differs from post-training compression, and practical inference still depends on compatible kernels and runtimes.
- Original title
- 面壁智能联合清华等开源中国首个基于华为昇腾训练的 1.58-bit 端侧大模型 BitCPM-CANN
- Source
- IT之家 · www.ithome.com
- Topic
- Compact models
- Source month
- 2026-05
This is a concise EasyHub summary of the linked source, not the full report or original reporting. Availability and preview conditions are described in the summary and original.
Summary page published · Editorial information