---
title: "What Is An AI Model? — Thoughtful Comics"
description: "The trained engine behind every AI tool, what shapes it, and how to pick one for your own work."
source: "https://thoughtfulrobots.ai/thoughtful-comics/what-is-an-ai-model"
---

# What Is An AI Model?

The trained engine behind every AI tool, what shapes it, and how to pick one for your own work.

![A four-panel comic. Panel 1. The reader: “I used the same prompt in two tools. Why was one answer much better?” The robot: “They may use different AI models, each trained and designed with different strengths.” Panel 2. The reader: “What is inside a model?” The robot: “It contains many learned parameters that shape how input becomes a predicted output.” Panel 3. The reader: “So the model controls everything the app can do?” The robot: “No. The product can add tools, context, permissions, and guardrails around the model.” Panel 4. The reader: “I will test models on my real tasks instead of choosing only by popularity.” The robot: “Exactly. The best model is the one that meets your actual requirements.” Glossary. AI model: A trained mathematical system that predicts or generates outputs from inputs. Training: The process of adjusting a model by learning patterns from examples. Parameter: A numerical setting inside a model that changes as it learns. Inference: The moment a trained model uses what it learned to produce an answer.](/images/comics/what-is-an-ai-model/strip.webp)

Click to enlarge ⤢

## The takeaways

- 01 The model is the trained engine behind an AI capability.

- 02 Training adjusts a model's internal parameters so it can recognise and generate patterns.

- 03 Model capability and product capability are related but not identical.

- 04 Evaluate a model using representative work, not a single impressive demo.

- Reader I used the same prompt in two tools. Why was one answer much better? Robot They may use different AI models, each trained and designed with different strengths. Takeaway: The model is the trained engine behind an AI capability.

- Reader What is inside a model? Robot It contains many learned parameters that shape how input becomes a predicted output. Takeaway: Training adjusts a model's internal parameters so it can recognise and generate patterns.

- Reader So the model controls everything the app can do? Robot No. The product can add tools, context, permissions, and guardrails around the model. Takeaway: Model capability and product capability are related but not identical.

- Reader I will test models on my real tasks instead of choosing only by popularity. Robot Exactly. The best model is the one that meets your actual requirements. Takeaway: Evaluate a model using representative work, not a single impressive demo.