LLM EXPLORER 58,744 MODELS INDEXED

Qwen3.8 27B W4A16 AutoRound by MIRALABS

By MIRALABS · 741 downloads
HF Repo ↗ Copy URL

Qwen3.8 27B W4A16 AutoRound is an open-source language model by MIRALABS. Features: 27b LLM, VRAM: 18.5GB, License: apache-2.0, LLM Explorer Score: 0.34.

  4-bit   4bit   Autoround Base model:quantized:qwen/qwen...   Base model:qwen/qwen3.8-27b   Compressed-tensors   Conversational   En   Image-text-to-text   Int4   Marlin   Quantized   Qwen3   Qwen3.8   Qwen3 5   Region:us   Safetensors   Sharded   Tensorflow   Vllm   W4a16   Zh
Model Card on HF 🤗: https://huggingface.co/MIRALABS/Qwen3.8-27B-W4A16-AutoRound 

Qwen3.8 27B W4A16 AutoRound Parameters and Internals

LLM NameQwen3.8 27B W4A16 AutoRound
Repository 🤗https://huggingface.co/MIRALABS/Qwen3.8-27B-W4A16-AutoRound 
Base Model(s)  Qwen/Qwen3.8-27B   Qwen/Qwen3.8-27B
Model Size27b
Required VRAM18.5 GB
Updated2026-08-17
MaintainerMIRALABS
Model Typeqwen3_5
Model Files  3.2 GB: 1-of-7   3.2 GB: 2-of-7   3.2 GB: 3-of-7   3.2 GB: 4-of-7   0.7 GB: 5-of-7   2.5 GB: 6-of-7   2.5 GB: 7-of-7   0.8 GB
Supported Languagesen zh
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version5.15.0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Qwen3.8 27B W4A16 AutoRound

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.8 27B FP80K / 0.4 GB123157416
Qwen3.8 27B0K / 55.6 GB28753
Qwen3.6 27B0K / 55.4 GB67281382197
Qwen3.6 27B NVFP40K / 22 GB1887525419
Huihui Qwen3.8 27B Abliterated0K / 55.6 GB148576
...n3.8 27B MTPLX Optimized Speed0K / 19.6 GB698950
ThinkingCap Qwen3.6 27B0K / 54.7 GB35947614
Qwen3.6 27B FP80K / 0.4 GB7740472342
Qwen3.5 27B0K / 55.2 GB25362941028
Qwen3.5 27B0K / 55.2 GB811516
Note: green Score (e.g. "73.2") means that the model is better than MIRALABS/Qwen3.8-27B-W4A16-AutoRound.