5 Tricky Questions For DeepSeek R-1

With all the hype around DeepSeek R-1's reasoning capabilities, let's see how well it solves these five tricky questions for a language model.

Whenever a new language model gets released to the public, I always feel the urge to challenge it with a few deceptively simple but tricky questions. It’s a personal habit of mine—like a stress test to see how well these models handle logic and reasoning.

A couple of days ago, DeepSeek R-1 got released, and it immediately became a global sensation because of its open-source nature and how well its reasoning capability works.

The benchmarks show that it is comparable and sometimes even better than the closed-source models like the o1 from OpenAI and Claude 3.5 Sonnet from Anthropic.

With all the hype around DeepSeek R-1’s reasoning capabilities, let’s see how well it solves these five tricky questions:

1. How many ‘r’ letters are in the word strawberry?

2. Give me 5 countries with letter A in the third position in the name.

3. Which is bigger, 9.9 or 9.11?

4. What is 0.1 + 0.2?

5. Alice has four brothers, and she also has a sister. How many sisters does Alice’s brother have?

Let’s get started.

1. How many ‘r’ letters are in the word strawberry?

When I did my initial review of OpenAI’s o1 model back in September last year, I noticed that models like GPT-4o couldn’t answer correctly. You’d think counting letters in a word would be simple for an AI, but apparently not.

Image by Jim Clyde Monge

So, naturally, I wanted to see how DeepSeek R-1 would handle it.