What a Model Is

A "model" is a trained mathematical system.

Core concept:
- Billions of numbers (parameters) that encode patterns
- Trained on massive datasets (text, images, etc.)
- Different versions have different capabilities

Common models:
- GPT-4 — OpenAI's flagship, strong at reasoning
- Claude 3.5 — Anthropic's latest, strong at nuance
- Gemini Ultra — Google's most capable model
- LLaMA — Meta's open-source models

Why versions matter:
- Newer ≠ always better for your task
- Different models have different strengths
- Some are faster, some are more capable
- Cost varies significantly by model