You can deploy this model and start earning money today!

majentik/
Qwen3.8-27B-MLX-4bit

Name: Qwen3.8-27B-MLX-4bit
Availability: InStock

Views

No views yet

Qwen3.8-27B-MLX-4bit

MLX 4bit (affine, group size 32) quantized variant of Qwen/Qwen3.8-27B (text tower quantized; vision tower and projector retained in BF16) for Apple silicon via mlx-lm.

Provenance

Source: Qwen/Qwen3.8-27B @ revision 1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 (Apache-2.0 (upstream LICENSE)).
Quantized with mlx_lm.convert (mlx-lm 0.31.3): affine, 4-bit, group size 32.

Smoke gate

Before upload this pack passed a deterministic coherence gate: greedy 48-token chat generation loaded through mlx_lm.load, judged for emptiness, repetition loops, multi-script gibberish, and special-token debris. Verdict: ok.

Usage

bash

pip install mlx-lm
mlx_lm.generate --model majentik/Qwen3.8-27B-MLX-4bit --prompt "Hello"

Evaluation

Benchmark	Score
arc_easy_acc	0.5900
hellaswag_acc	0.5150

License

Apache-2.0 — see the upstream LICENSE file in Qwen/Qwen3.8-27B.

Available tiers

Repo	Quant
majentik/Qwen3.8-27B-MLX-2bit	2bit
majentik/Qwen3.8-27B-MLX-3bit	3bit
majentik/Qwen3.8-27B-MLX-4bit	4bit
majentik/Qwen3.8-27B-MLX-5bit	5bit
majentik/Qwen3.8-27B-MLX-6bit	6bit
majentik/Qwen3.8-27B-MLX-8bit	8bit
majentik/Qwen3.8-27B-MLX-MXFP4	MXFP4