What are the hardware requirements for DeepSeek V3.1 Base BF16?

DeepSeek V3.1 Base BF16 requires approximately 358.8 GB of VRAM and supports a context window of 160K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed DeepSeek V3.1 Base BF16 and how large is it?

DeepSeek V3.1 Base BF16 is developed by unsloth, a model with 684.5b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate DeepSeek V3.1 Base BF16?

DeepSeek V3.1 Base BF16 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

DeepSeek V3.1 Base BF16 by unsloth — VRAM 358.8GB, 160K context

Name: DeepSeek V3.1 Base BF16
Author: unsloth

DeepSeek V3.1 Base BF16 is an open-source language model by unsloth. Features: 684.5b LLM, VRAM: 358.8GB, Context: 160K, License: mit, LLM Explorer Score: 0.2.

Arxiv:2412.19437 Base model:deepseek-ai/deepsee... Base model:finetune:deepseek-a... Conversational Custom code Deepseek Deepseek v3 Endpoints compatible Region:us Safetensors Sharded Tensorflow Unsloth

Model Card on HF 🤗: https://huggingface.co/unsloth/DeepSeek-V3.1-Base-BF16

DeepSeek V3.1 Base BF16 Parameters and Internals

LLM Name	DeepSeek V3.1 Base BF16
Repository 🤗	https://huggingface.co/unsloth/DeepSeek-V3.1-Base-BF16
Base Model(s)	deepseek-ai/DeepSeek-V3.1-Base deepseek-ai/DeepSeek-V3.1-Base
Model Size	684.5b
Required VRAM	358.8 GB
Updated	2026-08-06
Maintainer	unsloth
Model Type	deepseek_v3
Model Files	8.6 GB: 1-of-163 8.6 GB: 2-of-163 8.6 GB: 3-of-163 8.6 GB: 4-of-163 8.6 GB: 5-of-163 8.7 GB: 6-of-163 8.6 GB: 7-of-163 8.6 GB: 8-of-163 8.6 GB: 9-of-163 8.6 GB: 10-of-163 8.6 GB: 11-of-163 2.6 GB: 12-of-163 8.6 GB: 13-of-163 8.6 GB: 14-of-163 8.6 GB: 15-of-163 8.6 GB: 16-of-163 8.6 GB: 17-of-163 8.6 GB: 18-of-163 8.6 GB: 19-of-163 8.6 GB: 20-of-163 8.6 GB: 21-of-163 8.6 GB: 22-of-163 8.6 GB: 23-of-163 8.6 GB: 24-of-163 8.6 GB: 25-of-163 8.6 GB: 26-of-163 8.6 GB: 27-of-163 8.6 GB: 28-of-163 8.6 GB: 29-of-163 8.6 GB: 30-of-163 8.6 GB: 31-of-163 8.6 GB: 32-of-163 8.6 GB: 33-of-163 3.5 GB: 34-of-163 8.6 GB: 35-of-163 8.6 GB: 36-of-163 8.6 GB: 37-of-163 8.6 GB: 38-of-163 8.6 GB: 39-of-163 8.6 GB: 40-of-163 8.6 GB: 41-of-163 8.6 GB: 42-of-163 8.6 GB: 43-of-163
Model Architecture	DeepseekV3ForCausalLM
License	mit
Context Length	163840
Model Max Length	163840
Transformers Version	4.55.2
Vocabulary Size	129280
Torch Data Type	bfloat16

Rank the DeepSeek V3.1 Base BF16 capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to DeepSeek V3.1 Base BF16

Best Alternatives	Context / RAM	Downloads	Likes
DeepSeek R1 0528	160K / 180.4 GB	244573	2457
DeepSeek V3	160K / 171.8 GB	1155192	4102
DeepSeek R1	160K / 180.4 GB	8739585	13461
DeepSeek V3 Base	160K / 171.8 GB	34207	1700
DeepSeek V3.0324	160K / 184.7 GB	835522	3148
DeepSeek V3	160K / 171.8 GB	20952	13
DeepSeek V3.1 Terminus	160K / 171.8 GB	15	4
DeepSeek V3.1	160K / 180.4 GB	307929	826
DeepSeek V3.1	160K / 176.1 GB	16	3
DeepSeek R1 Block INT8	160K / 176.1 GB	171	50

Note: green Score (e.g. "73.2") means that the model is better than unsloth/DeepSeek-V3.1-Base-BF16.