Written by Jakub Rusinowski · Last updated March 1, 2024
A research paradigm shift by Microsoft. BitNet b1.58 replaces 16-bit weights with ternary {-1, 0, 1} weights. This eliminates Matrix Multiplication (MatMul) in favor of simple addition, offering extreme speed and efficiency on CPUs.
| Licence | What it permits | Applies to |
|---|---|---|
MIT | Commercial use permitted Commercial use permitted. No usage restrictions beyond attribution. | BitNet b1.58 3B |
| BitNet b1.58 3B | Min 3 GB VRAM · 1.58-bit · 2,048 ctx · ollama run hf.co/1bitLLM/bitnet_b1_58-3B |
The cheapest GPU that runs BitNet b1.58 locally (min 3 GB VRAM) is the Intel Arc B570 (10 GB).
Install Ollama then run: ollama run hf.co/1bitLLM/bitnet_b1_58-3B
Minimum VRAM: 3 GB. For best results use Q4_K_M quantization.
BitNet b1.58 needs about 3 GB VRAM at 1.58-bit quantization for its smallest variant. Variants: BitNet b1.58 3B (3 GB, 1.58-bit). On Apple Silicon, unified memory counts toward this requirement.
Yes — BitNet b1.58 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at 1.58-bit.
Q4_K_M is the best balance of quality and VRAM for BitNet b1.58 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run hf.co/1bitLLM/bitnet_b1_58-3B. This downloads BitNet b1.58 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.