Deepseek AI better than ChatGPT Strawberry? | Best FREE Reasoning AI App

The Revolution of AI Reasoning

When ChatGPT launched their O1 model (codename Strawberry), there was huge hype around it because this was the first model that could think before speaking. Giving it this ability unlocked PhD level intelligence. Now there's a new AI that gives us this ability for free - DeepSeek AI.

Understanding the Five Levels of AI

OpenAI released five levels of artificial intelligence showing progression from chatbots to AGI (artificial general intelligence). We started with chatbots for conversational language, but the next level is "reasoners" - AI that can think at human level problem solving. The O1 model took us to this next level with a reasoning step where it thinks about its response before speaking.

How AI Reasoning Actually Works

Original chatbots would immediately give an answer based on input - you input your prompt and get the output. With O1 reasoning, the AI goes through a chain of thought before giving their answer. It takes the input, does some reasoning, creates output, then uses that output as input for a second step of reasoning, and continues this process through multiple steps before arriving at the final answer.

The Impressive Results of Reasoning AI

The results from adding this reasoning step were remarkable. O1 scored in the 89th percentile in competitive programming questions, placed in the top 500 students in the Math Olympiad, and exceeds PhD level accuracy on benchmarks in physics, biology, and chemistry. The longer they gave the model to think with more reasoning steps, the more accurate the answers became - creating a new way to scale intelligence after training.

DeepSeek R1: The Free Alternative

Unfortunately, ChatGPT O1 was behind a paywall requiring a subscription. Now DeepSeek, which is open source, has provided this reasoning capability to anyone for free. You can ask 50 reasoning questions per day, which is competitive with the paid ChatGPT version that also limits you to 50 reasoning questions.

DeepSeek offers this free as a marketing tactic to attract developers and get them to taste how good the system is, hoping they'll use their paid API for app development.

The Transparency Advantage

What makes DeepSeek superior to ChatGPT O1 is transparency. You can actually see everything the AI is thinking, whereas ChatGPT only shows a truncated, summarized version of the thinking process. ChatGPT even bans people who try to investigate further and prompt the system to reveal its reasoning steps.

Live Demo: Watching AI Think

In a live comparison using the famous "how many R's in strawberry" prompt, both systems arrived at the correct answer of 3 R's. However, DeepSeek's transparent thinking process was fascinating to observe. The AI showed uncertainty, self-doubt, and multiple verification steps - almost like watching a human work through a problem.

The AI went through spelling the word multiple times: "S-T-R-A-W-B-E-R-R-Y", counted the R's (one after T, two consecutive R's after E), double-checked dictionaries, and even considered etymology before confidently concluding there are 3 R's in strawberry.

Unexpected Thinking Patterns

What was most surprising was which questions made DeepSeek think longest. The trolley dilemma (ethical problem about sacrificing one life to save five) took 25 seconds. Achieving peace in the Middle East took only 9 seconds. But the hypothetical question "would you rather fight one horse-sized duck or 100 duck-sized horses" took an incredible 205 seconds of deep analysis.

The AI considered factors like body weight, potential for injury, elements of surprise, attack patterns, and strategic advantages - treating this playful question with serious analytical rigor.

How to Use DeepSeek Effectively

My recommendation is to supplement your regular AI usage (GPT-4, Claude, Perplexity) by also running complex questions through DeepSeek. Use your 50 free daily reasoning prompts for problems requiring deep analysis. The transparency lets you understand how AI thinks and develops reasoning - adding a fascinating extra layer to your AI interactions.

This reasoning mode is particularly valuable for complex problems requiring multi-step analysis. You can access DeepSeek completely free at deepseek.com and experience PhD-level AI reasoning without the paywall.