LLM EXPLORER 59,278 MODELS INDEXED

DeepSeek V4 Flash 0731 Exl3 Hybrid 2.72bpw by anoane

By anoane · 150 downloads
HF Repo ↗ Copy URL

DeepSeek V4 Flash 0731 Exl3 Hybrid 2.72bpw is an open-source language model by anoane. Features: 52.2b LLM, VRAM: 104.5GB, Context: 1024K, License: mit.

Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-...   Conversational   Deepseek v4   Exl3   Exllamav3   Mixture-of-experts   Moe   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/anoane/DeepSeek-V4-Flash-0731-exl3-hybrid-2.72bpw 

DeepSeek V4 Flash 0731 Exl3 Hybrid 2.72bpw Parameters and Internals

LLM NameDeepSeek V4 Flash 0731 Exl3 Hybrid 2.72bpw
Repository 🤗https://huggingface.co/anoane/DeepSeek-V4-Flash-0731-exl3-hybrid-2.72bpw 
Base Model(s)  DeepSeek V4 Flash 0731   deepseek-ai/DeepSeek-V4-Flash-0731
Model Size52.2b
Required VRAM104.5 GB
Updated2026-08-23
Maintaineranoane
Model Typedeepseek_v4
Model Files  6.1 GB: 1-of-16   7.1 GB: 2-of-16   6.8 GB: 3-of-16   6.8 GB: 4-of-16   6.7 GB: 5-of-16   6.8 GB: 6-of-16   6.8 GB: 7-of-16   6.9 GB: 8-of-16   6.8 GB: 9-of-16   6.7 GB: 10-of-16   6.8 GB: 11-of-16   6.7 GB: 12-of-16   6.7 GB: 13-of-16   6.7 GB: 14-of-16   8.3 GB: 15-of-16   1.8 GB: 16-of-16   5.2 GB
Model ArchitectureDeepseekV4ForCausalLM
Licensemit
Context Length1048576
Model Max Length1048576
Transformers Version4.57.1
Tokenizer ClassPreTrainedTokenizerFast
Beginning of Sentence Token<｜begin▁of▁sentence｜>
End of Sentence Token<｜end▁of▁sentence｜>
Vocabulary Size129280
Torch Data Typebfloat16