LLM EXPLORER 63,069 MODELS INDEXED

Swift 1.5 Qwen3.8 27B GPTQ Int4 Baked V1 Embed Int8 by Wrapzii

By Wrapzii · 39 downloads
HF Repo ↗ Copy URL

Swift 1.5 Qwen3.8 27B GPTQ Int4 Baked V1 Embed Int8 is an open-source language model by Wrapzii. Features: 27b LLM, VRAM: 16.2GB, License: other, Quantized, LLM Explorer Score: 0.3.

  4-bit Base model:quantized:ukisai/sw... Base model:ukisai/swift-1.5-qw...   Conversational   Gptq   Image-text-to-text   Int4   Int8-embedding   Intel-xpu   Mtp   Quantized   Qwen3 5   Qwen3 8   Reasoning   Region:us   Safetensors   Sharded   Tensorflow   Vllm
Model Card on HF 🤗: https://huggingface.co/Wrapzii/Swift-1.5-Qwen3.8-27b-GPTQ-Int4-baked-v1-embed-int8 

Swift 1.5 Qwen3.8 27B GPTQ Int4 Baked V1 Embed Int8 Parameters and Internals

LLM NameSwift 1.5 Qwen3.8 27B GPTQ Int4 Baked V1 Embed Int8
Repository 🤗https://huggingface.co/Wrapzii/Swift-1.5-Qwen3.8-27b-GPTQ-Int4-baked-v1-embed-int8 
Base Model(s)  ukisai/Swift-1.5-Qwen3.8-27b-W4A16-AutoRound   ukisai/Swift-1.5-Qwen3.8-27b-W4A16-AutoRound
Model Size27b
Required VRAM16.2 GB
Updated2026-10-05
MaintainerWrapzii
Model Typeqwen3_5
Model Files  0.0 GB   5.0 GB: 2-of-5   5.0 GB: 3-of-5   5.0 GB: 4-of-5   1.2 GB: 5-of-5   0.9 GB   1.3 GB   0.0 GB
GPTQ QuantizationYes
Quantization Typegptq
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseother
Model Max Length262144
Transformers Version5.14.1
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Swift 1.5 Qwen3.8 27B GPTQ Int4 Baked V1 Embed Int8

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.6 27B GPTQ Pro 4bit0K / 18.8 GB41540739
Qwen3.8 27B GPTQ INT4 FP8KV0K / 20.2 GB64882
Qwen3.5 27B GPTQ Int40K / 30.1 GB10538157
...en3.8 27B W4A16 AutoRound GPTQ0K / 19.2 GB37485
Qwen3.8 27B GPTQ 4bit0K / 21 GB24335
Qwen3.8 27B GPTQ Int40K / 20.2 GB22673
...7B GPTQ Int4 Sym G128 MTP BF160K / 19.6 GB1753
Qwen3.8 27B GPTQ INT4 G1280K / 19.6 GB20323
Qwen3.8 27B GPTQ Int40K / 19.6 GB8811
...7B GPTQ Int4 Sym G128 MTP BF160K / 19.6 GB3221
Note: green Score (e.g. "73.2") means that the model is better than Wrapzii/Swift-1.5-Qwen3.8-27b-GPTQ-Int4-baked-v1-embed-int8.