Qwopus3.6-27B-v2 MLX 8bit

Near-lossless 8-bit MLX of Jackrong/Qwopus3.6-27B-v2. Group size 64.

Use

pip install mlx-lm
mlx_lm.generate --model zaydiscold/Qwopus3.6-27B-v2-MLX-8bit \
  --prompt "Explain quantum entanglement" --max-tokens 200 --temp 0.8

Author's temperature recommendation: 0.75–1.0.

The MLX ladder

Variant	Repo
bf16	`Qwopus3.6-27B-v2-MLX-bf16`
8bit (this repo)	this
4bit	`Qwopus3.6-27B-v2-MLX-4bit`

Credits

Source: Jackrong/Qwopus3.6-27B-v2 by Jackrong
MLX 8-bit conversion by zaydiscold

Mix brought by the NOTORIOUS MLX.

Downloads last month: 58

Safetensors

Model size

8B params

Tensor type

BF16

U32

MLX

Hardware compatibility

8-bit

Inference Providers NEW

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

Model tree for zaydiscold/Qwopus3.6-27B-v2-MLX-8bit

Base model

Jackrong/Qwopus3.6-27B-v2

Quantized

(22)

this model

Collection including zaydiscold/Qwopus3.6-27B-v2-MLX-8bit

Qwopus3.6-27B-v2 MLX

Collection

MLX quant suite of Jackrong/Qwopus3.6-27B-v2 for Apple Silicon. Mix brought by the NOTORIOUS MLX. • 16 items • Updated 1 day ago