Hy4-flash MLX 4.6-bit

A 125.20 GB quant of Hy4-flash from Nace.AI, 4.6 bits per weight.

On a fixed 100-question LiveCodeBench release_v6 subset with an 8,192-token cap, it solved 39/100 questions, the same as the 4-bit pack at a smaller size.

Hy4-flash quantization quality across 37 measured variants

37 measured Hy4-flash variants on the same 11 texts, 5,621 scored tokens. Chart sizes use text-weight payload, including required Engram tables, excluding vision/MTP. Complete weight-file sizes and exact scores are in the comparison data.

Run on macOS

Download the model and run it with the bundled loader:

hf download nace-ai/Hy4-flash-MLX-4.6bit --local-dir Hy4-flash-MLX-4.6bit
cd Hy4-flash-MLX-4.6bit
pip install -r requirements-macos.txt
python run_mlx.py --prompt 'What is the capital of France?' --max-tokens 32

License

HY Community License.

Downloads last month
-
Safetensors
Model size
149B params
Tensor type
U32
·
F32
·
BF16
·
I32
·
F16
·
MLX
Hardware compatibility
Log In to add your hardware

4-bit

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

Model tree for nace-ai/Hy4-flash-MLX-4.6bit

Quantized
(40)
this model

Collection including nace-ai/Hy4-flash-MLX-4.6bit