LLM EXPLORER 60,897 MODELS INDEXED

Qwen3.8 27B PARO MXFP4 by Launch80

By Launch80 · 192 downloads
HF Repo ↗ Copy URL

Qwen3.8 27B PARO MXFP4 is an open-source language model by Launch80. Features: 27b LLM, VRAM: 19.1GB, License: apache-2.0.

  Arxiv:2511.10645   4-bit Base model:quantized:qwen/qwen...   Base model:qwen/qwen3.8-27b   Conversational   Image-text-to-text   Mxfp4   Paroquant   Paroquant mxfp4   Quantized   Qwen3 5   Rdna4   Region:us   Rocm   Safetensors   Vllm   W4a8
Model Card on HF 🤗: https://huggingface.co/Launch80/Qwen3.8-27B-PARO-MXFP4 

Qwen3.8 27B PARO MXFP4 Parameters and Internals

LLM NameQwen3.8 27B PARO MXFP4
Repository 🤗https://huggingface.co/Launch80/Qwen3.8-27B-PARO-MXFP4 
Base Model(s)  Qwen/Qwen3.8-27B   Qwen/Qwen3.8-27B
Model Size27b
Required VRAM19.1 GB
Updated2026-09-12
MaintainerLaunch80
Model Typeqwen3_5
Model Files  19.1 GB
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 PARO MXFP4

Best Alternatives
Context / RAM
Downloads
Likes
Agens Pilot1024K / 54.7 GB1271
Qwen3.6 27B0K / 55.4 GB67281382197
Qwen3.8 27B MLX0K / 16 GB141118
Qwen3.8 27B0K / 55.6 GB272721
Qwen3.8 27B FP80K / 0.4 GB123157416
Qwen3.8 27B0K / 55.6 GB28753
Qwen3.6 27B NVFP40K / 22 GB1887525419
Qwen3.8 27B NVFP40K / 22 GB1919574
Qwen3.6 27B FP80K / 0.4 GB7740472342
Cortex Mini 1 Preview0K / 55.6 GB1252
Note: green Score (e.g. "73.2") means that the model is better than Launch80/Qwen3.8-27B-PARO-MXFP4.