Comparing DeepSeek-R1 Models: 32B vs 70B vs R1

Author(s): Lorentz Yeung

Originally published on Towards AI.

DeepSeek has made waves in the AI world. They offer multiple models at the same time, so which one should we choose?

This member-only story is on us. Upgrade to access all of Medium.

DeepSeek, a Chinese AI company founded in 2023 by Liang Wenfeng, has made significant strides in artificial intelligence with its release of the DeepSeek-R1 model. This open-source language model has garnered attention for its performance in reasoning tasks, rivaling models like OpenAI’s o1. Notably, DeepSeek achieved this with a fraction of the resources typically required, highlighting China’s rapid advancement in AI technology.

In my exploration of DeepSeek’s offerings, I evaluated two of their models: DeepSeek-R1-Distill-Qwen-32B and DeepSeek-R1-Distill-Llama-70B. Both models are available for download from Ollama official website through this link. (deepseek-r1) For example, if you would like to download the 32B model, which is 20GB in size, you can just follow the instruction in this page.

Pip install Ollama first, then run this command “ollama run deepseek-r1:32b” in your terminal. Picuture by the author.

For more details about DeepSeek’s research and architecture, you can check out the official DeepSeek website (deepseek.com).

I ran these models on my i7–14700KF 3.4GHz processor, 32GB RAM, and an NVIDIA RTX 4090 GPU, using WSL2.

The 32B model ran smoothly without any system modifications.The 70B model required setting memory to 24GB, which I managed using psutil to monitor RAM… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI