MarkChen1214/cohere-transcribe-03-2026-MLX-Mixed-3bit4bit
Automatic Speech Recognition • 0.3B • Updated • 69 • 4
ASR Model Compression
Compressed ASR for Apple Silicon: Cohere Transcribe and NVIDIA Nemotron in MLX/CoreML, including multilingual and Arabic checkpoints.
-
-
MarkChen1214/cohere-transcribe-03-2026-MLX-Mixed-2bit3bit4bit
Automatic Speech Recognition • 0.2B • Updated • 59 • 1 -
MarkChen1214/cohere-transcribe-arabic-07-2026-MLX-Mixed-2bit3bit4bit
Automatic Speech Recognition • 0.2B • Updated • 147 • 1 -
MarkChen1214/nemotron-3.5-asr-streaming-0.6b-MLX-Mixed-6bit8bit
Automatic Speech Recognition • 0.2B • Updated • 113 -
MarkChen1214/parakeet-tdt-0.6b-v3-MLX-Mixed-4bit8bit
Automatic Speech Recognition • 0.1B • Updated • 73
MarkChen1214/Qwen3-ASR-1.7B-MLX-Mixed-4bit6bit8bit
Automatic Speech Recognition • 0.4B • Updated • 70Note Qwen3-ASR 1.7B mixed 4/6/8-bit MLX model; 1.30 GB with 5.51% macro WER on the 100-sample comparison.
-
MarkChen1214/parakeet-tdt-0.6b-v2-MLX-Mixed-3bit4bitBF16
Automatic Speech Recognition • 0.1B • Updated • 65 -
MarkChen1214/parakeet-unified-en-0.6b-MLX-Mixed-3bit4bit6bit
Automatic Speech Recognition • 88M • Updated • 50 -
MarkChen1214/cohere-transcribe-03-2026-CoreML-INT4
Automatic Speech Recognition • Updated • 7
MarkChen1214/Breeze-ASR-25-MLX-5bit
Automatic Speech Recognition • 0.3B • Updated • 24Note Breeze-ASR-25 uniform 5-bit MLX model; 1.07 GB with 10.30% CER on all 4,976 Common Voice 16.1 zh-TW test samples and 17.7x RTFx on the complete stable-power timing lane.