DeepSeek R1 vs. ChatGPT: The AI Model Shaking Up the Industry (and It’s Free!)

DeepSeek is Dominating Headlines

At this point, it's impossible to ignore DeepSeek. The app has skyrocketed to the number one spot on the App Store charts, sparking discussions across the tech world and even causing waves in the stock market. Some users are already claiming it’s a better alternative to ChatGPT. But is it really that groundbreaking? In this blog post, I’ll break down exactly what’s happening, why DeepSeek R1 is making headlines, and whether the hype is justified.

What is DeepSeek R1?

DeepSeek is an AI company out of China that recently launched a new open-source AI reasoning model called R1. 

This model is designed for multi-step thinking and problem-solving rather than just predicting text like a glorified autocomplete.

What Makes It So Special?

What makes R1 special is that it matches the performance of OpenAI’s cutting-edge GPT-4 Turbo reasoning model, which is locked behind a $20/month subscription. 

DeepSeek, on the other hand, is completely free and open-source. This is a massive deal because not only can you use the web interface, but you can also run this locally on your computer without sending your data to the cloud.

This is What Caused the Panic

The stock market sold off, and people started panicking when it was revealed that this model allegedly only cost under $6 million to build and train. However, this number is just for one training, so the total cost is likely higher. Still, it’s significantly less than what OpenAI has spent on training their models.

Significant Cost Savings

For consumers and enterprises, the API usage is significantly cheaper. R1 costs $219 per million output tokens, while OpenAI’s GPT-4 Turbo model costs $60 per million output tokens. 

That’s a fraction of the price and a massive win for both consumers and enterprises building apps with this model.

How is This Different from ChatGPT?

DeepSeek matches the performance of GPT-4 Turbo in math, coding, and reasoning tasks, but what makes it unique is its approach. 

It uses a combination of supervised fine-tuning and reinforcement learning (RL) to handle complex reasoning tasks, whereas most AI models rely more heavily on supervised fine-tuning with less emphasis on RL.

Get DeepSeek on iPhone

To use DeepSeek, search for it in the App Store. It’s currently number one in every chart. 

The app is frequently updated, so version numbers may vary, but once you download it, you’ll need to make an account. Signing in with Apple or Google is the quickest way to get started.

Don’t Worry About iOS App Stealing Data

iOS and macOS apps run in a sandboxed environment, meaning they cannot access other app data or system files without explicit permission. 

If you're worried about the app stealing contacts or messages, that won't happen unless you explicitly allow it.

Using DeepSeek R1 on iPhone

Once inside the app, you need to enable DeepSeek R1, as it’s off by default. You can also enable web search, which is something even OpenAI’s GPT-4 Turbo model cannot do.

Chain of Thought in Action

DeepSeek R1 displays its reasoning process before giving an answer. This is called chain-of-thought reasoning. When solving a problem, it writes out its thought process step by step, making it much more transparent than ChatGPT.

Reinforcement Learning

DeepSeek uses RL, meaning that if it makes a mistake, it learns from itself before providing a final answer. If you tell it that it's wrong, it will analyze the mistake and correct itself.

Apple Intelligence Fails the Test (As Expected)

Apple Intelligence simply isn't in the same league. Running the same test, Apple’s AI attempted to start a workout instead of answering the question. It was a funny but clear failure.

ChatGPT App on iPhone

In the ChatGPT app, GPT-4 Turbo answered correctly but did not show its reasoning like DeepSeek did. It simply provided the answer without walking through the problem.

DeepSeek R1 vs ChatGPT-4 Turbo Performance

Surprisingly, DeepSeek R1 finished solving a problem before ChatGPT-4 Turbo did. This was unexpected given OpenAI’s dominance, but it shows how competitive DeepSeek is.

Writing Code with DeepSeek vs ChatGPT

DeepSeek allows users to run HTML code directly in the interface, something ChatGPT does not. However, ChatGPT produced a slightly better final result when generating a 3D ball floating in space.

This Was Eye-Opening – DeepSeek for the Win (Writing Test)

For a writing task, DeepSeek produced a more humorous and engaging script than ChatGPT. It included a joke about “Tim Apple” and was overall more personable and creative.

How to Install DeepSeek Locally on Mac (or Any AI Model)

To run DeepSeek locally, use LM Studio. Download it, search for DeepSeek R1, and select a smaller, distilled model version to run on consumer-grade hardware. Running the model locally means no data is sent to the cloud.

Privacy Concerns – Should You Be Worried?

DeepSeek's privacy policy states that user data is stored in secure servers in China, collecting keystroke patterns, IP addresses, and more. However, OpenAI collects a similar amount of data. If you're concerned, you can run R1 locally or use US-based providers like Together AI to keep data off Chinese servers.

Overall Thoughts on DeepSeek – Big Deal or Overhyped?

DeepSeek is one of the most powerful models on the market and completely free. While ChatGPT still holds an edge with its voice mode and custom GPTs, DeepSeek is a game-changer for enterprises and individuals looking for a high-performance, cost-effective AI alternative.