cff-version: 1.2.0
message: "If you use WAL-Ternary-8B or its direct-packed runtime, please cite this software."
title: "WAL-Ternary-8B: a 2.898-BPW Qwen3-8B with direct-packed CPU/CUDA inference"
type: software
authors:
  - family-names: Aubakirov
    given-names: Arman
version: 0.2.0
date-released: 2026-08-07
url: "https://github.com/AubakirovArman/WAL-Ternary-8B"
repository-code: "https://github.com/AubakirovArman/WAL-Ternary-8B"
keywords:
  - ternary quantization
  - low-bit inference
  - Qwen3
  - CUDA
  - CPU inference