LLM EXPLORER 58,425 MODELS INDEXED

Qwen3.8 2.4T A95B NVFP4 Pruned75 by mgoin

By mgoin · 127 downloads
HF Repo ↗ Copy URL

Qwen3.8 2.4T A95B NVFP4 Pruned75 is an open-source language model by mgoin. Features: 348.9b LLM, VRAM: 436.9GB, Context: 256K, License: other, MoE, LLM Explorer Score: 0.3.

  8-bit   Conversational   Endpoints compatible   Modelopt   Moe   Qwen3 5 moe text   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/mgoin/Qwen3.8-2.4T-A95B-NVFP4-pruned75 

Qwen3.8 2.4T A95B NVFP4 Pruned75 Parameters and Internals

LLM NameQwen3.8 2.4T A95B NVFP4 Pruned75
Repository 🤗https://huggingface.co/mgoin/Qwen3.8-2.4T-A95B-NVFP4-pruned75 
Model Size348.9b
Required VRAM436.9 GB
Updated2026-08-14
Maintainermgoin
Model Typeqwen3_5_moe_text
Model Files  15.9 GB: 1-of-7   15.9 GB: 2-of-7   15.8 GB: 3-of-7   15.9 GB: 4-of-7   15.9 GB: 5-of-7   15.8 GB: 6-of-7   2.3 GB: 7-of-7   10.0 GB: 1-of-34   10.0 GB: 2-of-34   10.0 GB: 3-of-34   10.0 GB: 4-of-34   10.0 GB: 5-of-34   10.0 GB: 6-of-34   10.0 GB: 7-of-34   10.0 GB: 8-of-34   10.0 GB: 9-of-34   10.0 GB: 10-of-34   10.0 GB: 11-of-34   10.0 GB: 12-of-34   10.0 GB: 13-of-34   10.0 GB: 14-of-34   10.0 GB: 15-of-34   10.0 GB: 16-of-34   10.0 GB: 17-of-34   10.0 GB: 18-of-34   10.0 GB: 19-of-34   10.0 GB: 20-of-34   10.0 GB: 21-of-34   10.0 GB: 22-of-34   10.0 GB: 23-of-34   10.0 GB: 24-of-34   10.0 GB: 25-of-34   10.0 GB: 26-of-34   10.0 GB: 27-of-34   10.0 GB: 28-of-34   10.0 GB: 29-of-34   10.0 GB: 30-of-34   10.0 GB: 31-of-34   10.0 GB: 32-of-34   10.0 GB: 33-of-34   9.4 GB: 34-of-34
Model ArchitectureQwen3_5MoeForCausalLM
Licenseother
Context Length262144
Model Max Length262144
Transformers Version4.57.3
Vocabulary Size248320