LLM EXPLORER 62,233 MODELS INDEXED

Qwen3.8 27B INT8 W8A8 Imatrix Heretic MTP by RukaRat

By RukaRat · 43 downloads
HF Repo ↗ Copy URL

Qwen3.8 27B INT8 W8A8 Imatrix Heretic MTP is an open-source language model by RukaRat. Features: 27b LLM, VRAM: 30.4GB, License: apache-2.0, LLM Explorer Score: 0.28.

  8-bit   Abliterated Base model:quantized:qwen/qwen...   Base model:qwen/qwen3.8-27b   Compressed-tensors   Conversational   Endpoints compatible   Heretic   Image-text-to-text   Int8   Llmcompressor   Mtp   Multimodal   Qwen3 5   Region:us   Safetensors   Sglang   Sharded   Speculative-decoding   Tensorflow   Uncensored   Vllm   W8a8
Model Card on HF 🤗: https://huggingface.co/RukaRat/Qwen3.8-27B-INT8-W8A8-imatrix-heretic-MTP 

Qwen3.8 27B INT8 W8A8 Imatrix Heretic MTP Parameters and Internals

LLM NameQwen3.8 27B INT8 W8A8 Imatrix Heretic MTP
Repository 🤗https://huggingface.co/RukaRat/Qwen3.8-27B-INT8-W8A8-imatrix-heretic-MTP 
Base Model(s)  Qwen/Qwen3.8-27B   Qwen/Qwen3.8-27B
Model Size27b
Required VRAM30.4 GB
Updated2026-09-26
MaintainerRukaRat
Model Typeqwen3_5
Model Files  20.0 GB: 1-of-2   10.4 GB: 2-of-2   0.8 GB
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version5.14.1
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Qwen3.8 27B INT8 W8A8 Imatrix Heretic MTP

Best Alternatives
Context / RAM
Downloads
Likes
...wopus3.8 27B Flash Mlx OQ4e 1M1024K / 16.9 GB2111
Agens Pilot1024K / 54.7 GB1721
Qwen3.6 27B0K / 55.4 GB67281382197
Qwen3.8 27B0K / 55.6 GB28753
Qwen3.8 27B FP80K / 0.4 GB123157416
Qwen3.8 27B0K / 55.6 GB272721
Qwen3.8 27B MLX0K / 16 GB141118
Qwen3.6 27B NVFP40K / 22 GB1887525419
Qwen3.5 27B0K / 55.2 GB25362941028
Qwen3.6 27B FP80K / 0.4 GB7740472342
Note: green Score (e.g. "73.2") means that the model is better than RukaRat/Qwen3.8-27B-INT8-W8A8-imatrix-heretic-MTP.