Skip to content

AI Model

An AI model is a trained computational system that transforms inputs into predictions, classifications, generated content, or other outputs.

An AI model is a trained computational system that transforms inputs into predictions, classifications, generated outputs, or other results.

The model represents learned behaviour produced through a training process.

What a Model Can Receive

Depending on the model, inputs can include:

  • Text
  • Images
  • Audio
  • Video
  • Numerical information
  • Structured data

The supported inputs and outputs are commonly called modalities.

Model vs. Application

The model is one part of an AI application.

A production system may also contain:

User
 ↓
Application
 ↓
Prompt
 ↓
Retrieval
 ↓
AI model
 ↓
Tools
 ↓
Validation
 ↓
User

Changing the model can affect the application, but the surrounding architecture remains important.

Model Choice

Different models may prioritize different combinations of:

  • Capability
  • Speed
  • Context Window
  • Cost
  • Tool support
  • Input modalities
  • Output limits

Current model providers therefore typically expose several models rather than one universal option.

OpenAI's current model catalogue, for example, documents model capabilities, context limits, supported tools, pricing, and other characteristics. OpenAI model catalogue

Why It Matters

“Use AI” is not a complete technical requirement.

A system needs a model suited to the task.

A simple classification workflow may not need the same model as:

  • Long-document analysis
  • Complex coding
  • Image interpretation
  • Agentic work
  • High-volume extraction

Common Problems

Common problems include:

  • Always choosing the largest available model
  • Choosing entirely on cost
  • No evaluation before switching models
  • Assuming newer always means better for the specific use case
  • Application tightly coupled to one model's quirks
  • No fallback if a model is retired

How to Improve

Define the task and evaluation criteria first.

Then compare models using AI Evaluation rather than anecdotes.

Document model dependencies as part of Technical Documentation and Dependency management.