LLM EXPLORER 58,531 MODELS INDEXED

Qwen3.8 27B MTPLX 4bit by johninthepool

By johninthepool · 272 downloads
HF Repo ↗ Copy URL

Qwen3.8 27B MTPLX 4bit is an open-source language model by johninthepool. Features: 27b LLM, VRAM: 15.1GB, License: apache-2.0, Quantized, LLM Explorer Score: 0.32.

  4-bit   4bit   Apple-silicon Base model:quantized:qwen/qwen...   Base model:qwen/qwen3.8-27b   Conversational   Mlx   Mtplx   Multi-token-prediction   Quantized   Qwen3.5   Qwen3 5   Region:us   Safetensors   Sharded   Speculative-decoding   Tensorflow
Model Card on HF 🤗: https://huggingface.co/johninthepool/Qwen3.8-27B-MTPLX-4bit 

Qwen3.8 27B MTPLX 4bit Parameters and Internals

LLM NameQwen3.8 27B MTPLX 4bit
Repository 🤗https://huggingface.co/johninthepool/Qwen3.8-27B-MTPLX-4bit 
Base Model(s)  Qwen/Qwen3.8-27B   Qwen/Qwen3.8-27B
Model Size27b
Required VRAM15.1 GB
Updated2026-08-15
Maintainerjohninthepool
Model Typeqwen3_5
Model Files  5.3 GB: 1-of-3   5.3 GB: 2-of-3   4.5 GB: 3-of-3   0.8 GB
Quantization Type4bit
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 MTPLX 4bit

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.6 27B OptiQ 4bit256K / 18.7 GB1306559
Bonsai 27B Mlx 1bit0K / 5.1 GB809309205
Qwen3.8 27B 4bit0K / 16 GB605527
Ternary Bonsai 27B Mlx 2bit0K / 8.5 GB809024168
Qwen3.8 27B MLX 4bit0K / 16 GB1971210
Qwen3.6 27B MLX 8bit0K / 29.4 GB87914514
...X Qwen3.8 27B MLX AXQ 6bit MTP0K / 19 GB10885
Qwen3.6 27B MLX 6bit0K / 22.7 GB8129772
Qwen3.6 27B MLX 4bit0K / 16 GB84686116
...X Qwen3.8 27B MLX AXQ 4bit MTP0K / 16.4 GB5542
Note: green Score (e.g. "73.2") means that the model is better than johninthepool/Qwen3.8-27B-MTPLX-4bit.