ModelBest releases BitCPM-CANN, 1.58-bit ternary LLM trained on Ascend NPUs

Hugging Face Models

BitCPM-CANN achieves 1.58-bit ternary quantization with 6x less VRAM than BF16 while retaining most capability. The model family includes sizes from 2B to 8B and is fully open-source on HuggingFace.

See 3 more sources

ModelBest just dropped BitCPM-CANN, the first 1.58-bit ternary LLM trained end-to-end on Ascend...17 days agoHasan Toor

OpenBMB presents the model BitCPM-CANN 1.58 bit19 days agoIllustrious-Swim9663Discuss

BitCPM-CANN: Native 1.58-Bit Large Language Model Training on Ascend NPU17 days agoAaaaaaaaaeeeeeDiscuss

Hugging Face Models·openbmb

26 days ago