LLM EXPLORER 58,608 MODELS INDEXED

Qwen3.8 P44w75 16.8B Unrepaired by logic65

By logic65 · 929 downloads
HF Repo ↗ Copy URL

Qwen3.8 P44w75 16.8B Unrepaired is an open-source language model by logic65. Features: 27b LLM, VRAM: 2.4GB, Context: 256K, License: apache-2.0, LLM Explorer Score: 0.34.

Base model:quantized:qwen/qwen... Base model:qwen/qwen3.8-27b-fp...   Conversational   Depth-pruning   Endpoints compatible   Gated-deltanet   Gguf   Pruning   Qwen3 5   Region:us   Safetensors   Width-pruning   Zero-training
Model Card on HF 🤗: https://huggingface.co/logic65/Qwen3.8-p44w75-16.8B-unrepaired 

Qwen3.8 P44w75 16.8B Unrepaired Parameters and Internals

LLM NameQwen3.8 P44w75 16.8B Unrepaired
Repository 🤗https://huggingface.co/logic65/Qwen3.8-p44w75-16.8B-unrepaired 
Base Model(s)  Qwen/Qwen3.8-27B-FP8   Qwen/Qwen3.8-27B-FP8
Model Size27b
Required VRAM2.4 GB
Updated2026-08-16
Maintainerlogic65
Model Typeqwen3_5
Model Files  2.5 GB   4.0 GB   4.0 GB   3.9 GB   4.0 GB   3.9 GB   4.0 GB   4.0 GB   2.4 GB
Model ArchitectureQwen3_5ForCausalLM
Licenseapache-2.0
Context Length262144
Model Max Length262144
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size248320
Torch Data Typebfloat16
Errorsreplace

Best Alternatives to Qwen3.8 P44w75 16.8B Unrepaired

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.6 27B OBLITERATED256K / 53.6 GB15137184
...2 Abliterated OptiQ 3.7bpw Mlx256K / 13.4 GB672714
...2 Abliterated OptiQ 3.7bpw Mlx256K / 13.4 GB591914
Qwen3.6 27B Esper4256K / 55.6 GB29219
Ornstein 3.6 27B RYS256K / 56.8 GB3971
Qwen3.6 Solidity 27B256K / 53.8 GB876
Qwen3.6 27B Tachibana Agent256K / 55.6 GB607
Carnice 27B256K / 53.7 GB2736
...wopus3.5 27B V3 FP8 Vllm Ready256K / 29.4 GB5017
Qwopus3.5 27B V3 FP8256K / 29.4 GB516
Note: green Score (e.g. "73.2") means that the model is better than logic65/Qwen3.8-p44w75-16.8B-unrepaired.