PTENES
Skip to content
MODULE 1 - FUNDAMENTALS

🤖 LLM Basics and How They Work

Large Language Models are neural networks trained on billions of texts to understand and generate language. Understanding how they work is essential for creating effective prompts.

100B+
Parameters
1T+
Training Tokens
~4
Chars/Token
🧠

What Is an LLM?

Understanding the technology behind generative AI

An LLM (Large Language Model) is an AI model trained on billions of words from internet text, books, articles, and other sources. Through this massive training, it learns language patterns, factual knowledge, and reasoning capabilities.

⚙️ How an LLM Works

1

Training

Billions of texts

2

Patterns

Identifies rules

3

Prompt

You ask

4

Generation

Token by token

Token Prediction

The model predicts the most likely next word based on everything that came before, creating coherent text.

Attention

Mechanism that allows the model to “pay attention” to relevant parts of the context when generating each word.

📊 Main Characteristics

  • • It is not conscious: It’s a sophisticated mathematical tool; it doesn’t think or feel like humans do.
  • • Probability-based: Chooses words based on which one is most likely to come next in context.
  • • No persistent memory: Each conversation is isolated—it doesn't “remember” previous conversations.
  • • Cutoff date: Knowledge limited to the training date (doesn't know recent events).
🏆

Popular Models in 2025

Get to know the leading LLMs on the market

🟠 Claude (Anthropic)

  • ✓ Excellent at complex reasoning
  • ✓ Strong at following long instructions
  • ✓ 200K context tokens
  • ✓ Focus on safety and ethics

🟢 GPT-4 (OpenAI)

  • ✓ Very versatile and creative
  • ✓ Broad general knowledge
  • ✓ Strong at coding and analysis
  • ✓ Integration with plugins

🔵 Gemini (Google)

  • ✓ Integrated with Google
  • ✓ Strong at research and facts
  • ✓ Native multimodal
  • ✓ 1M tokens of context
💬

How the LLM Responds

Understand the text generation process

When you ask "What is the capital of Brazil?", the LLM doesn't "know" the answer — it recognizes patterns and generates the most likely continuation.

❌ What the LLM DOESN'T do

  • ✗ Query a database
  • ✗ Access the internet in real time
  • ✗ "Remember" previous conversations
  • ✗ Reason like a human

✓ What the LLM DOES

  • ✓ Analyzes your prompt word by word
  • ✓ Identifies training patterns
  • ✓ Generates the most likely tokens in sequence
  • ✓ Continues until the answer is complete

💡 Crucial Insight

That’s why clear prompts help: you guide the model toward the right patterns!

The model doesn’t “know” things—it recognizes patterns and generates likely responses. The more specific your prompt, the easier it is for the model to find the right patterns.

⭐

Practical Tips

How to use LLMs effectively

💡 LLMs are better at tasks with lots of examples in their training—common tasks tend to work better.
💡 The clearer and more specific your prompt, the better the response—avoid ambiguity.
💡 LLMs can “hallucinate”—make up information that seems real. Always verify important facts.
💡 Use LLMs as assistants, not definitive sources of truth—they’re tools, not oracles.
💡 Different models have different strengths—experiment to find the best one for your task.
⚠️

Common Errors

What to avoid when using LLMs

❌ Treating the LLM as omniscient

Example: Asking About Events After the Training Date

✓ Solution: Provide context if the information is recent or highly specific. Check the model's cutoff date.

❌ Assuming the LLM "understands"

Example: Expecting the Model to Have Human Common Sense

✓ Solution: Be explicit in your instructions; don't assume anything. Detail what you want.

🚀 Next Step

Now that you understand how LLMs work, learn about Tokens and Context Window to understand the limits of what you can do!

Go to Tokens