jjang-ai/vmlx

vMLX - Cont Batch, Prefix, Paged, KV Cache Quant, VL - Powers MLX Studio. Image gen/edit, OpenAI/Anth

42
/ 100
Emerging

Implements adaptive mixed-precision quantization (JANG 2-bit) with a 5-layer cache hierarchy combining memory-aware prefix caching, paged KV blocks, and persistent disk storage with LRU eviction. Built on MLX's Metal GPU acceleration for Apple Silicon, it exposes OpenAI/Anthropic-compatible REST APIs while supporting tool calling, reasoning modes, and multimodal inference across LLMs, VLMs, MoE models, and diffusion-based image generation.

Available on PyPI.

Maintenance 13 / 25
Adoption 6 / 25
Maturity 18 / 25
Community 5 / 25

How are scores calculated?

Stars

15

Forks

1

Language

Python

License

Apache-2.0

Last pushed

Mar 18, 2026

Commits (30d)

0

Dependencies

19

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/vector-db/jjang-ai/vmlx"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.