CITATION.cff

# Citation File Format (CFF) v1.2.0
# Docs: https://citation-file-format.github.io/
cff-version: 1.2.0
title: "mlxcel: High-performance LLM/VLM/VLA inference on Apple Silicon and CUDA GPUs"
message: "If you use this software, please cite it as below."
type: software
authors:
  - name: "Lablup Inc."
    website: "https://lablup.com"
repository-code: "https://github.com/lablup/mlxcel"
url: "https://github.com/lablup/mlxcel"
abstract: >-
  mlxcel is a Rust inference runtime and OpenAI-compatible model server for
  MLX-format checkpoints. It targets Apple Silicon (primary) and Linux/CUDA
  (secondary), provides continuous batching, prompt-prefix caching, speculative
  decoding, KV-cache compression, and tensor/pipeline parallelism for
  selected model families. Model coverage tracks the mlx-lm and mlx-vlm
  projects where practical.
keywords:
  - mlx
  - llm
  - vlm
  - vla
  - inference
  - apple-silicon
  - cuda
  - rust
license: Apache-2.0
version: "0.0.27"
date-released: "2026-05-18"