LLM EXPLORER 67,181 MODELS INDEXED

Qwen3.8 27B Randean MXFP4 by flechin

By flechin · 59 downloads
HF Repo ↗ Copy URL

Qwen3.8 27B Randean MXFP4 is an open-source language model by flechin. Features: 27b LLM, VRAM: 19.8GB, License: apache-2.0, Quantized, LLM Explorer Score: 0.34.

  8-bit   Anti-sycophancy   Awq Base model:quantized:qwen/qwen...   Base model:qwen/qwen3.8-27b   Conversational   De-hedged   Endpoints compatible   Epistemology   Image-text-to-text   Model-index   Mxfp4   Orpo   Qlora   Quantized   Quark   Qwen3 5   Reality-aligned   Reasoning   Region:us   Rocm   Safetensors
Model Card on HF 🤗: https://huggingface.co/flechin/Qwen3.8-27B-Randean-MXFP4 

Qwen3.8 27B Randean MXFP4 Parameters and Internals

LLM NameQwen3.8 27B Randean MXFP4
Repository 🤗https://huggingface.co/flechin/Qwen3.8-27B-Randean-MXFP4 
Base Model(s)  Qwen/Qwen3.8-27B   amd/Qwen3.8-27B-Quark-AWQ-MXFP4   Qwen/Qwen3.8-27B   amd/Qwen3.8-27B-Quark-AWQ-MXFP4
Model Size27b
Required VRAM19.8 GB
Updated2026-11-11
Maintainerflechin
Model Typeqwen3_5
Model Files  19.8 GB
AWQ QuantizationYes
Quantization Typeawq
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version5.2.0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Qwen3.8 27B Randean MXFP4

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.6 27B AWQ INT40K / 20.4 GB2563390103
Qwen3.6 27B AWQ0K / 22 GB66892424
Qwen3.6 27B AWQ 4bit0K / 18.7 GB2603423
Qwen3.6 27B AWQ BF16 INT40K / 28.3 GB41917141
Qwen3.8 27B AWQ INT40K / 21 GB2101532
...wen3.6 27B Abliterated AWQ MTP0K / 18.4 GB24497812
Qwen3.5 27B AWQ0K / 21.9 GB37092044
Qwen3.6 27B AWQ INT40K / 20.4 GB2785301
Qwen3.5 27B AWQ 4bit0K / 20 GB27906442
Qwen3.8 27B AWQ W4A16 ASYM0K / 0.8 GB232164
Note: green Score (e.g. "73.2") means that the model is better than flechin/Qwen3.8-27B-Randean-MXFP4.