MOSS-Music-8B-Instruct — MLX 4-bit

An MLX conversion of OpenMOSS-Team/MOSS-Music-8B-Instruct (Apache-2.0), quantized to 4-bit (group size 64) with the moss_music_mlx converter from the community MLX port of the MOSS-Music inference stack — the same converter and layout as the mlx-community/MOSS-Music-8B-Thinking-* quants. The language model, adapters, lm_head and token embeddings are quantized; the audio encoder is kept in bf16.

Weights are otherwise unmodified. All credit to the OpenMOSS team; see the base model card for capabilities, benchmarks and citation.

Converted for use in a local macOS music app; uploaded so its model catalog has a stable download source.

Downloads last month: 102

Safetensors

Model size

9B params

Tensor type

U32

BF16

MLX

Hardware compatibility

Quantized

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for rtikw/MOSS-Music-8B-Instruct-4bit

Base model

OpenMOSS-Team/MOSS-Music-8B-Instruct

Finetuned

(3)

this model