LLM EXPLORER 65,695 MODELS INDEXED

MiniMax M3.1 Flash EXL3 SAGE 4.26bpw by vcruz305

By vcruz305 · 67 downloads
HF Repo ↗ Copy URL

MiniMax M3.1 Flash EXL3 SAGE 4.26bpw is an open-source language model by vcruz305. Features: 117.7b LLM, VRAM: 225GB, License: other.

  Arxiv:2606.13392 Base model:minimaxai/minimax-m... Base model:quantized:minimaxai...   Conversational   Custom code   En   Exl3   Exllamav3   Image-text-to-text   Minimax m3 vl   Moe   Quantized   Region:us   Safetensors   Sage   Sharded   Tensorflow   Zh
Model Card on HF 🤗: https://huggingface.co/vcruz305/MiniMax-M3.1-Flash-EXL3-SAGE-4.26bpw 

MiniMax M3.1 Flash EXL3 SAGE 4.26bpw Parameters and Internals

LLM NameMiniMax M3.1 Flash EXL3 SAGE 4.26bpw
Repository 🤗https://huggingface.co/vcruz305/MiniMax-M3.1-Flash-EXL3-SAGE-4.26bpw 
Base Model(s)  MiniMaxAI/MiniMax-M3.1-Flash   MiniMaxAI/MiniMax-M3.1-Flash
Model Size117.7b
Required VRAM225 GB
Updated2026-10-27
Maintainervcruz305
Model Typeminimax_m3_vl
Model Files  7.5 GB: 1-of-30   7.5 GB: 2-of-30   7.5 GB: 3-of-30   7.5 GB: 4-of-30   7.5 GB: 5-of-30   7.5 GB: 6-of-30   7.5 GB: 7-of-30   7.5 GB: 8-of-30   7.5 GB: 9-of-30   7.5 GB: 10-of-30   7.5 GB: 11-of-30   7.5 GB: 12-of-30   7.5 GB: 13-of-30   7.5 GB: 14-of-30   7.5 GB: 15-of-30   7.5 GB: 16-of-30   7.5 GB: 17-of-30   7.5 GB: 18-of-30   7.5 GB: 19-of-30   7.5 GB: 20-of-30   7.5 GB: 21-of-30   7.5 GB: 22-of-30   7.5 GB: 23-of-30   7.5 GB: 24-of-30   7.5 GB: 25-of-30   7.5 GB: 26-of-30   7.5 GB: 27-of-30   7.5 GB: 28-of-30   7.5 GB: 29-of-30   7.5 GB: 30-of-30
Supported Languagesen zh
Model ArchitectureMiniMaxM3SparseForConditionalGeneration
Licenseother
Model Max Length40960000
Transformers Version4.52.4
Tokenizer ClassPreTrainedTokenizerFast
Padding Token]!p~[
Torch Data Typebfloat16