A Gentle Introduction to Reasoning Models

In partnership with

Hire Ava, the Industry-Leading AI BDR

Your BDR team is wasting time on things AI can automate. Our AI BDR Ava automates your entire outbound demand generation so you can get leads delivered to your inbox on autopilot.

She operates within the Artisan platform, which consolidates every tool you need for outbound:

• 300M+ High-Quality B2B Prospects, including E-Commerce and Local Business Leads

• Automated Lead Enrichment With 10+ Data Sources

• Full Email Deliverability Management

• Multi-Channel Outreach Across Email & LinkedIn

• Human-Level Personalization

Book a demo to see what Ava can do.

Hello everyone and welcome to my newsletter where I discuss real-world skills needed for the top data jobs. 👏

This week I’m discussing the rise of the reasoning models.  👀

 Not a subscriber? Join the informed. Over 200K people read my content monthly.

 Thank you. 🎉

I just finished learning about the fundamentals of large language models and now I need to start learning about reasoning models? I can’t keep up and I do this for a living. The truth is, no one can. If you’re interested in a career with Ai, and most of us are in some form, then you’re going to need to learn the foundations and then focus on a niche. The pace of change in this space is insane. 🤯

I often make the mistake of assuming my readers know what Ai is. Most do but let’s just high level it before we dig in. This is a general high-level look and not meant to be the gospel. I just want to provide you with a general idea of where we are.

What is artificial intelligence? AI is the ability of a computer or machine to do tasks that usually require human intelligence. Not a great definition but that’s what ChatGPT gave me.

Next, all real-world artificial intelligence is machine learning. What is machine learning? Here’s the answer that been around for decades and it’s the best one. Machine learning is a way for computers to learn from data and improve their performance without being explicitly programmed.

Now let’s visualize the Ai hierarchy and find out where large language models fit in. The acronym ANN is an artificial neural network. The acronym DLM is a deep learning model. The acronym LLM is a large language model. 

At the top we have Ai. Inside of Ai we have machine learning models. Within machine learning there are two core types of models. There are traditional models and artificial neural networks. Inside of artificial neural networks are deep learning models. Inside of deep learning models are large language models. Inside of large language models are reasoning models.

Large language models are generative Ai models. Most modern generative AI relies on deep learning, but generative methods existed before it and can exist outside of it. Whew. 😴

Here’s a great prompt that will help you understand the line separating a deep learning model versus an artificial neural network.

what is the difference between an artificial neural network and a deep learning model

📷 Time to get focused on reasoning models. 

A reasoning model is a type of large language model (LLM) that can perform complex reasoning tasks. Instead of quickly generating output based solely on a statistical guess of what the next word(token) should be in an answer, as an LLM typically does, a reasoning model will take time to break a question down into individual steps and work through a chain of thought process to come up with a more accurate answer.  Unlike general-purpose LLMs which might generate direct answers, reasoning models are specifically trained to show their work and follow a more structured thought process.

Here’s a simple example between a vanilla LLM and a LLM using a reasoning model. If we were to ask ChatGPT what the capital of the US was, it would simply search through the data and find the response. Very similar to rote memorization for humans.

However, if we were to ask ChatGPT to calculate the time it would take us to travel from NYC to DC in a car if we were traveling at 80 mph, the model (nor the human) can no longer lookup the answer. The model (and the human) must perform some calcuations. A reasoning model is designed to simulate logical thinking, draw inferences, and solve problems using structured rules or symbolic logic.

Fact-based news without bias awaits. Make 1440 your choice today.

Overwhelmed by biased news? Cut through the clutter and get straight facts with your daily 1440 digest. From politics to sports, join millions who start their day informed.

Sign up now!

OpenAI debuted its first reasoning models, dubbed o1, in September 2024. In a blog post, the company explained that it used reinforcement learning (RL) techniques to train the reasoning model to handle complex tasks in mathematics, science, and coding. The model performed at the level of PhD students for physics, chemistry, and biology, while exceeding the ability of PhD students for math and coding.

Similar to how a human may think for a long time before responding to a difficult question, o1 uses a chain of thought when attempting to solve a problem, OpenAI said in a technical blog post. Through reinforcement learning, o1 learns to hone its chain of thought and refines the strategies it uses. It learns to recognize and correct its mistakes. It learns to break down tricky steps into simpler ones.

The model is designed to try a different approach when the current one isn’t working. This process dramatically improves the model’s ability to reason. Rather than just generating answers, today’s reasoning models can break down their analysis step-by-step. This allows the models to tackle ever more complex problems, guiding users to take meaningful action.

Chain of Thought is a technique where the model explains its reasoning in steps before giving an answer. 👀

How do reasoning models work? Reasoning in AI is depicted as a system typically made up of two core components. The first system is called knowledge based and the second one an inference engine. 

The knowledge base is the backbone of a reasoning system. It contains knowledge graphs, ontologies, semantic networks and other models of knowledge representation. These structured forms map real-world entities—such as concepts, domain-specific information, events, facts, objects, relationships, rules and situations—into a structure that the machine learning models can process and understand.

Acting as the brain of a reasoning system is the inference engine. It’s powered by trained machine learning models. The inference engine implements the needed logic and reasoning methods to analyze data from the knowledge base and reach a decision.

To illustrate how a reasoning system works, let’s take an autonomous robotic floor cleaner as an example. Its knowledge base can contain information about different kinds of floors and what type of cleaning they require. The robot’s machine learning models have also been trained to recognize and classify each floor type based on this knowledge base.

When deployed for cleaning, the robot receives and processes input data, including images and sensor data. Then, it draws upon its knowledge base and training and applies the appropriate reasoning technique to make a real-time decision on its cleaning action, such as vacuuming and mopping hardwood, tile and vinyl floors but only vacuuming carpeted floors.

When do we need a reasoning model? Reasoning models are designed to be good at complex tasks such as solving puzzles, advanced math problems, and challenging coding tasks. However, they are not necessary for simpler tasks like summarization, translation, or knowledge-based question answering.

In fact, using reasoning models for everything can be inefficient and expensive. For instance, reasoning models are typically more expensive to use, more verbose, and sometimes more prone to errors due to overthinking. The old adage goes, use the right tool (or type of LLM) for the task. 👏

Reasoning models represent a step forward in the maturity of large language models, enabling systems to not just recognize patterns but also to reason, think, and solve problems in a more human-like way.

Let’s highlight some core aspects of a reasoning model. 

• Reasoning models are deep learning models.

• Reasoning models are GenerativeAI models.

• Reasoning models are a type of large language model.

• Reasoning models use a chain of thought. This means explaining the steps behind each of their actions prior to giving an answer.

• Reasoning models are designed to try a different approach if the current one isn’t working.

• Reasoning models are often depicted with two core engines. One is called a knowledge base and the other an inference engine.

• Reasoning models are designed to excel at solving puzzles, advanced math problems, and challenging coding tasks.

• Reasoning models are not as good at simpler tasks like summarization, translation, or knowledge-based question answering.

SQL is the SINGLE most important skill for any data professional. I’ve curated the top questions for acing the SQL Server interview with my latest GPT release. I call the GPT SQL Server Hyper Focus on my platform called LogikBot.

Sounds more complicated than it is. It is simply a tool for preparing for the real-world SQL Server interview using ChatGPT. I’m an interview resource for Microsoft. I know what the questions are.

You have two purchasing options.

• As a stand alone GPT you add to ChatGPT. (You MUST have the PLUS versions of ChatGPT)

• With all my other courses including the recently published Fabric one.

🥳 The cost of the standalone GPT is $39. However, I’ll knock that down to $19 for the first 20 readers. Simply use this code at checkout: maynewsletter2025

Here’s what it looks like. 

The top 5 statements above are mine. They come from a document I added to my GPT. I provided the instructions and data to ChatGPT for the model. The model is then trained on my data.

The novice might be thinking… so what? Well, I’ve just given you an interactive way to pass the SQL Server part of most SQL Server related technical interviews. If you are completely new and don’t know what a technical interview is, here you go.

For those seeking any information technology role, the technical interview is where you lose and get the job. It’s stressful and unforgiving. You either do well or forget about that job. 🤯

There are no instructions. You learn the material how you want to. Simply ask ChatGPT to create a study guide and off you go. Using ChatGPT gives you a world of flexibility.

When you are using the GPT ask it anything you need clarification on. Once you understand what an index is, you might what to see the code behind it.

If you are serious about a job working with the Microsoft ecosystem, I guarantee this one GPT will be among the top purchases you make in your career.

Stay focused, stay hyper focused. 🎊