Today we’re covering the very buzzy topic of *Deepseek*, and distilling the news down to the facts.

But first, a little context: Deepseek is an open source AI model released by a Chinese venture & now research firm High-Flyer, sparking hyperbolic discussions of a full-blown AI arms race. But what about the tech itself? What do you, as a founder or corporation, really need to know? 

Should you be using GPT, Deepseek, or something else? Today we’ll answer that question.

First, let’s discuss cost. 

One of the groundbreaking features of Deepseek is that it is an open-source model, which means technically, it’s free. That’s right, you can download and run Deepseek’s R1 model on a local device, whereas OpenAI’s models are currently only available through API requests.

However there are some drawbacks to running Deepseek locally. First of all, the speed and dependability of your service will depend on how much compute, or hardware, you have on hand. Attempting to service thousands or even hundreds of customers by running Deepseek for free on your laptop just isn’t going to work - it would be way too slow. So if you’re not looking to give up your desk space to build a sudo-server room, then you’ll want to stick with APIs.

OpenAI APIs charge you for the compute and service, whereas Deepseek only charges you for compute. As you can see from the official pricing, that makes Deepseek much cheaper. 

And don’t worry, if you’re eager to switch-over, Deepseek is compatible with OpenAI’s SDK, meaning you won’t have to change your entire infrastructure to utilize this model instead.

What about model performance?

Deepseek published a paper exhibiting that their R1 model outperformed OpenAI’s most recent O1 model across the board.

This means, not only is it cheaper, but it’s also better. And because it’s open source, an entire global community of developers and students will have access to the model, and will help continually improve it. We commend Deepseek for choosing to open source their work - a move that will advance humanity’s collective knowledge of artificial intelligence and the bounds of its capabilities in the coming years! These are exciting times to be an innovator.

But although the power of open source is not to be diminished, it’s important to note that OpenAI has been around much longer, meaning they’ve had more time to develop best service practices. And due to Deepseek's meteoric popularity and influx of users, they are experiencing downtime issues. If you have users in need of constant AI servicing, consistency is definitely something you’ll want to consider.

OpenAI’s host of tools and integrations are also much more comprehensive, especially at the enterprise level. OpenAI offers audio and visual generative models, whereas Deepseek R1 is just a text model. However, because the SDKs are compatible, you can swap out your OpenAI text model for Deepseek’s, and keep the rest of your APIs.

What you choose depends on your specific use case, but here are some indicators that may help you choose the right AI solution for your business:

Choosing the right AI solution is only half the battle

Tech news headlines are constantly inundated with AI developments, which can feel overwhelming at times. And though it’s very important to stay current in order to supply your company with the best tools, the most important focus is still meeting customer needs and building efficient solutions to their problems. 

At Bedrock, we have an innovation forward approach to building optimized systems that ensure reliability, cost efficiency, and of course, ingenuity. From automating compliance checking processes with GPT to optimizing customer support mechanisms, we’ve consistently been integrating AI seamlessly into new and existing products at the edge of its ongoing evolution. 

We maintain that the most important focus for startups and corporations within the swirling noise of AI developments, is finding a source of expertise that you trust, and solving real problems in creative ways.

Thanks for tuning in! If you liked this article and found it useful, stick around for more tech news, case studies, and best practice tips coming soon.

Brinlee Kidd - Junior Developer @ Bedrock