LLM EXPLORER 61,405 MODELS INDEXED

Swift Qwen3.8 27B W4A16 AWQ by TheUnderscore

By TheUnderscore · 595 downloads
HF Repo ↗ Copy URL

Swift Qwen3.8 27B W4A16 AWQ is an open-source language model by TheUnderscore. Features: 27b LLM, VRAM: 18.8GB, License: other, Quantized, LLM Explorer Score: 0.35.

  Awq Base model:quantized:ukisai/sw... Base model:ukisai/swift-qwen3....   Compressed-tensors   Conversational   Endpoints compatible   Finetuned   Image-text-to-text   Int4   Lmdeploy   Quantized   Qwen   Qwen3.5   Qwen3.8   Qwen3 5   Region:us   Safetensors   Sharded   Tensorflow   W4a16
Model Card on HF 🤗: https://huggingface.co/TheUnderscore/Swift-Qwen3.8-27b-W4A16-AWQ 

Swift Qwen3.8 27B W4A16 AWQ Parameters and Internals

LLM NameSwift Qwen3.8 27B W4A16 AWQ
Repository 🤗https://huggingface.co/TheUnderscore/Swift-Qwen3.8-27b-W4A16-AWQ 
Base Model(s)  ukisai/Swift-Qwen3.8-27b   ukisai/Swift-Qwen3.8-27b
Model Size27b
Required VRAM18.8 GB
Updated2026-09-16
MaintainerTheUnderscore
Model Typeqwen3_5
Model Files  3.5 GB: 1-of-7   2.2 GB: 2-of-7   2.2 GB: 3-of-7   2.2 GB: 4-of-7   2.2 GB: 5-of-7   2.2 GB: 6-of-7   4.3 GB: 7-of-7   0.8 GB
AWQ QuantizationYes
Quantization Typeawq
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseother
Model Max Length262144
Transformers Version5.8.0.dev0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Swift Qwen3.8 27B W4A16 AWQ

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.6 27B AWQ INT40K / 20.4 GB2563390103
Qwen3.8 27B AWQ INT40K / 21 GB2101532
Qwen3.6 27B AWQ0K / 22 GB66892424
Qwen3.6 27B AWQ BF16 INT40K / 28.3 GB41917141
...wen3.6 27B Abliterated AWQ MTP0K / 18.4 GB24497812
Qwen3.8 27B AWQ 4bit0K / 27.7 GB118746
Qwen3.5 27B AWQ0K / 21.9 GB37092044
Qwen3.5 27B AWQ 4bit0K / 20 GB27906442
...wen3.8 27B Abliterated AWQ MTP0K / 19.1 GB44948
Qwen3.8 27B AWQ 5.0bpw0K / 17.3 GB144520
Note: green Score (e.g. "73.2") means that the model is better than TheUnderscore/Swift-Qwen3.8-27b-W4A16-AWQ.