kuotient
/

MOSS-Transcribe-Diarize-MLX-8bit

Automatic Speech Recognition

moss_transcribe_diarize

speaker-diarization

8-bit precision

Model card Files Files and versions

MOSS-Transcribe-Diarize — MLX 8-bit

8-bit MLX quantization of OpenMOSS-Team/MOSS-Transcribe-Diarize (0.9B end-to-end transcription + speaker diarization + timestamps).

Quantization: affine 8-bit, group size 64, text backbone only — the Whisper encoder and the VQ adaptor stay full precision (quantizing the encoder's positional embedding breaks the feature broadcast).
Quality: byte-identical transcripts to the bf16 reference across Korean, Korean↔English code-switching, and multi-speaker English in our checks.
Load: mlx-audio / mlx-audio-swift via fromModelDirectory.

Original model © the OpenMOSS / MOSI.AI team, Apache-2.0. This repository only re-hosts a quantized copy for reproducible on-device deployment.

Downloads last month: 67

Safetensors

Model size

0.6B params

Tensor type

BF16

·

U32

·

MLX

Hardware compatibility

Log In to add your hardware

8-bit

Model tree for kuotient/MOSS-Transcribe-Diarize-MLX-8bit

Base model

OpenMOSS-Team/MOSS-Transcribe-Diarize

Quantized

(11)

this model