# microsoft/BitNet

Official inference framework for 1-bit LLMs

Repository: https://github.com/microsoft/BitNet
Canonical: https://ross.abutalabs.com/products/bitnet
Language: C++
License: MIT
License Family: permissive
Last push: 2026-07-27T05:52:06+00:00

## Health v2 (maintenance only)
Score: 65/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 94, release rhythm 35, longevity 54
- inputs: {"age_days": 758, "days_push": 37, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 40203, forks 3723 (observed 2026-08-28T04:12:08.168167+00:00)

## What it is
bitnet.cpp is Microsoft's official inference framework for 1-bit large language models (BitNet b1.58), providing optimized CPU kernels with GPU support. It enables fast, energy-efficient inference of ternary-quantized LLMs and embedding models on edge devices.

## Use cases
- run 1-bit LLMs locally on CPU
- efficient inference of ternary-quantized models on edge devices
- convert and run BitNet embedding models with I2_S kernels
- speed up LLM inference without GPUs
- deploy small language models on resource-constrained hardware

## When to choose
- you need to run BitNet b1.58 or 1-bit embedding models efficiently on CPU
- you target edge or low-power hardware for LLM inference
- you want Microsoft's official, optimized kernels for ternary LLMs

## When to avoid
- you need to serve standard full-precision or mainstream quantized models (use llama.cpp or vLLM)
- you require broad model-family support beyond BitNet-style ternary models
- you need a managed API rather than a self-hosted inference engine

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, machine-learning, cli, gpu-computing
- domain: large-language-models, machine-learning, deep-learning, developer-tools
- platform: cpp, python, windows, cli
- tags: 1-bit-llm, ternary-quantization, bitnet, cpu-inference, edge-inference, quantization, embeddings, linux, macos, gpu

## Member repositories
- microsoft/BitNet (main) score 65

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:12:08.168167+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:24:04.911142+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/BitNet (fetched 2026-08-28T04:12:08.168167+00:00, sha 32a7c4bf724d)
- Data as of 2026-08-30T08:39:29.467469+00:00.