BRIEFING // 002
What is an
LLM?
The engine behind the chatbot. Understanding how it "thinks" helps you get better results.
KEY TAKEAWAYS
- 01It's a prediction engine, not a knowledge database.
- 02Key metrics: "Tokens" (cost) and "Context" (memory).
- 03New "Reasoning" models are solving the math/logic gap.
Definition
A Large Language Model (LLM) is a machine learning model trained on vast amounts of text. Its primary job is to predict the next word in a sequence. It is the engine behind tools like ChatGPT, Claude, and Gemini.
It doesn't "know" facts in the way a human does. Instead, it understands the statistical relationship between words. If you say "The sky is," it knows "blue" is a highly probable next word.
The Vocabulary
To use LLMs effectively, you need to understand three technical terms that affect daily use.
TOKEN
The unit of currency. Roughly 0.75 of a word. When you pay for API access, you pay per million tokens.
CONTEXT WINDOW
Short-term memory. If a model has a "128k context window," it can read a book's worth of text before it forgets the beginning.
TEMPERATURE
Creativity dial. Low temperature (0) is precise and factual. High temperature (1) is creative but risky.
How It Works
When you send a prompt, the model converts your text into numbers (vectors), processes them through layers of neural networks, and calculates the probability of every possible next word.
PROBABILISTIC VS DETERMINISTIC
Traditional software is Deterministic: Input A always equals Output B.
LLMs are Probabilistic: Input A likely equals Output B, but it might vary. This makes them great for creativity but dangerous for math without safeguards.
Limitations
Because they are prediction engines, LLMs can "hallucinate"—confidently stating facts that are wrong because they fit the pattern.
THE FIX
New "Reasoning Models" (like OpenAI's GPT-5.2 series) use a "chain of thought" to verify their own logic before answering, significantly reducing errors in math and coding.
NEXT STEPS
Want Help Getting Started?
Explore the full library or let us build a roadmap that fits your team.