LLM EXPLORER 58,608 MODELS INDEXED

DeepSeek V4 Pro 0813 Exl3 2.04bpw by bullerwins

By bullerwins · 104 downloads
HF Repo ↗ Copy URL

DeepSeek V4 Pro 0813 Exl3 2.04bpw is an open-source language model by bullerwins. Features: 217.4b LLM, VRAM: 293.6GB, Context: 1024K, License: mit.

  Arxiv:2606.19348 Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-...   Deepseek v4   Endpoints compatible   Exl3   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/bullerwins/DeepSeek-V4-Pro-0813-exl3-2.04bpw 

DeepSeek V4 Pro 0813 Exl3 2.04bpw Parameters and Internals

LLM NameDeepSeek V4 Pro 0813 Exl3 2.04bpw
Repository 🤗https://huggingface.co/bullerwins/DeepSeek-V4-Pro-0813-exl3-2.04bpw 
Base Model(s)  DeepSeek V4 Pro 0813   deepseek-ai/DeepSeek-V4-Pro-0813
Model Size217.4b
Required VRAM293.6 GB
Updated2026-08-17
Maintainerbullerwins
Model Typedeepseek_v4
Model Files  1.9 GB: 1-of-66   9.7 GB: 2-of-66   6.6 GB: 3-of-66   6.6 GB: 4-of-66   6.6 GB: 5-of-66   6.6 GB: 6-of-66   6.6 GB: 7-of-66   6.6 GB: 8-of-66   6.6 GB: 9-of-66   6.6 GB: 10-of-66   6.6 GB: 11-of-66   6.6 GB: 12-of-66   6.6 GB: 13-of-66   6.6 GB: 14-of-66   6.6 GB: 15-of-66   6.6 GB: 16-of-66   6.6 GB: 17-of-66   6.6 GB: 18-of-66   6.6 GB: 19-of-66   6.6 GB: 20-of-66   6.6 GB: 21-of-66   6.6 GB: 22-of-66   6.6 GB: 23-of-66   6.5 GB: 24-of-66   6.5 GB: 25-of-66   6.5 GB: 26-of-66   6.5 GB: 27-of-66   6.5 GB: 28-of-66   6.5 GB: 29-of-66   6.5 GB: 30-of-66   6.5 GB: 31-of-66   6.5 GB: 32-of-66   6.5 GB: 33-of-66   6.5 GB: 34-of-66   6.5 GB: 35-of-66   6.5 GB: 36-of-66   6.5 GB: 37-of-66   6.5 GB: 38-of-66   6.5 GB: 39-of-66   6.5 GB: 40-of-66   6.5 GB: 41-of-66   6.6 GB: 42-of-66   6.6 GB: 43-of-66   6.6 GB: 44-of-66   6.6 GB: 45-of-66
Model ArchitectureDeepseekV4ForCausalLM
Licensemit
Context Length1048576
Model Max Length1048576
Transformers Version4.57.1
Vocabulary Size129280
Torch Data Typebfloat16