microsoft/BitNet
Official inference framework for 1-bit LLMs
- Chart
- Not on a chart right now
- Stars
- 42.1k
What it is
bitnet.cpp is the official inference framework for 1-bit LLMs such as BitNet b1.58, providing optimized kernels for fast and lossless inference on CPUs and GPUs. It delivers significant speedups and energy reductions across x86 and ARM platforms.
Who it is for
AI developers and researchers looking to run quantized LLMs and embedding models efficiently on local or edge devices.
Why it is trending
It gained 367 stars today, reaching 41,865 total stars. Recent additions include 1-bit embedding models and VibeASR.cpp for real-time CPU speech recognition.
Main features
- Optimized 1.58-bit inference kernels for x86/ARM CPUs and GPUs
- Runs 100B BitNet models on a single CPU with reduced energy use
- Supports BitNet-b1.58-2B-4T and 1-bit embedding models (0.6B, 270M)
- Includes I2_S, TL1, and TL2 quantization kernels
Summarized on November 18, 2026 from the README and repository data. It can be wrong; check the source before relying on it. Report a wrong summary