LLM EXPLORER 57,252 MODELS INDEXED

DeepSeek V3.1 Int4 AutoRound by Intel

By Intel · 7 downloads
HF Repo ↗ Copy URL

DeepSeek V3.1 Int4 AutoRound is an open-source language model by Intel. Features: 2b LLM, VRAM: 225GB, Context: 160K, LLM Explorer Score: 0.2.

  Arxiv:2309.05516   4-bit   Auto-round Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-...   Conversational   Custom code   Deepseek v3   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/Intel/DeepSeek-V3.1-int4-AutoRound 

DeepSeek V3.1 Int4 AutoRound Parameters and Internals

LLM NameDeepSeek V3.1 Int4 AutoRound
Repository 🤗https://huggingface.co/Intel/DeepSeek-V3.1-int4-AutoRound 
Base Model(s)  deepseek-ai/DeepSeek-V3.1   deepseek-ai/DeepSeek-V3.1
Model Size2b
Required VRAM225 GB
Updated2026-01-23
MaintainerIntel
Model Typedeepseek_v3
Model Files  5.0 GB: 1-of-71   5.0 GB: 2-of-71   5.0 GB: 3-of-71   5.0 GB: 4-of-71   5.0 GB: 5-of-71   5.0 GB: 6-of-71   5.0 GB: 7-of-71   5.0 GB: 8-of-71   5.0 GB: 9-of-71   5.0 GB: 10-of-71   5.0 GB: 11-of-71   5.0 GB: 12-of-71   5.0 GB: 13-of-71   5.0 GB: 14-of-71   5.0 GB: 15-of-71   5.0 GB: 16-of-71   5.0 GB: 17-of-71   5.0 GB: 18-of-71   5.0 GB: 19-of-71   5.0 GB: 20-of-71   5.0 GB: 21-of-71   5.0 GB: 22-of-71   5.0 GB: 23-of-71   5.0 GB: 24-of-71   5.0 GB: 25-of-71   5.0 GB: 26-of-71   5.0 GB: 27-of-71   5.0 GB: 28-of-71   5.0 GB: 29-of-71   5.0 GB: 30-of-71   5.0 GB: 31-of-71   5.0 GB: 32-of-71   5.0 GB: 33-of-71   5.0 GB: 34-of-71   5.0 GB: 35-of-71   5.0 GB: 36-of-71   5.0 GB: 37-of-71   5.0 GB: 38-of-71   5.0 GB: 39-of-71   5.0 GB: 40-of-71   5.0 GB: 41-of-71   5.0 GB: 42-of-71   5.0 GB: 43-of-71   5.0 GB: 44-of-71   5.0 GB: 45-of-71
Model ArchitectureDeepseekV3ForCausalLM
Context Length163840
Model Max Length163840
Transformers Version4.56.0.dev0
Vocabulary Size129280
Torch Data Typebfloat16

Best Alternatives to DeepSeek V3.1 Int4 AutoRound

Best Alternatives
Context / RAM
Downloads
Likes
... Terminus Int4 Mixed AutoRound160K / 220 GB174
Note: green Score (e.g. "73.2") means that the model is better than Intel/DeepSeek-V3.1-int4-AutoRound.