LLM EXPLORER 59,071 MODELS INDEXED

Qwen3.8 27B 3bit MLX by lukaskremla

By lukaskremla · 648 downloads
HF Repo ↗ Copy URL

Qwen3.8 27B 3bit MLX is an open-source language model by lukaskremla. Features: 27b LLM, VRAM: 12.8GB, License: apache-2.0, Quantized, LLM Explorer Score: 0.32.

  3-bit   3bit   Affine Base model:quantized:qwen/qwen...   Base model:qwen/qwen3.8-27b   Bf16-vision-tower   Conversational   Group-size-64   Image-text-to-text   Long-context   Mlx   Mlx-vlm   Multilingual   Multimodal   Quantized   Qwen   Qwen3.8   Qwen3.8-27b   Qwen3 5   Reasoning   Region:us   Rtn   Safetensors   Sharded   Tensorflow   Tool-use   Video   Vision   Vision-language   Weight-only-quantization
Model Card on HF 🤗: https://huggingface.co/lukaskremla/Qwen3.8-27B-3bit-MLX 

Qwen3.8 27B 3bit MLX Parameters and Internals

LLM NameQwen3.8 27B 3bit MLX
Repository 🤗https://huggingface.co/lukaskremla/Qwen3.8-27B-3bit-MLX 
Model NameQwen3.8-27B 3-bit MLX
Base Model(s)  Qwen/Qwen3.8-27B   Qwen/Qwen3.8-27B
Model Size27b
Required VRAM12.8 GB
Updated2026-08-20
Maintainerlukaskremla
Model Typeqwen3_5
Model Files  5.4 GB: 1-of-3   5.4 GB: 2-of-3   2.0 GB: 3-of-3
Quantization Type3bit
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version5.8.0.dev0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Qwen3.8 27B 3bit MLX

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.6 27B OptiQ 4bit256K / 18.7 GB1306559
Bonsai 27B Mlx 1bit0K / 5.1 GB809309205
Ternary Bonsai 27B Mlx 2bit0K / 8.5 GB809024168
Qwen3.8 27B 8bit0K / 29.4 GB1389233
Qwen3.8 27B 4bit0K / 16 GB605527
Qwen3.8 27B MLX 4bit0K / 16 GB1971210
Qwen3.8 27B MLX 8bit0K / 29.4 GB248117
Qwen3.6 27B MLX 8bit0K / 29.4 GB87914514
Qwen3.8 27B OptiQ 4bit0K / 19.4 GB31899
Qwen3.6 27B MLX 6bit0K / 22.7 GB8129772
Note: green Score (e.g. "73.2") means that the model is better than lukaskremla/Qwen3.8-27B-3bit-MLX.