What is an LLM?

What is an LLM?

A Large Language Model (LLM) is a type of deep learning model trained on massive text datasets to understand and generate human language. It learns statistical patterns across billions of text examples to predict what text should come next in a sequence.

Core Concept

LLMs are trained with one simple objective: predict the next token given a sequence of tokens.

From this simple task, they develop emergent abilities: reasoning, coding, translation, summarization, and more.

Architecture

All modern LLMs are based on the Transformer architecture:

textCopy
Input Text → Tokenizer → Token Embeddings
                              ↓
                    [Transformer Blocks × N]
                    - Multi-head self-attention
                    - Feed-forward network
                    - Layer normalization
                              ↓
                    Output logits → Next token

Training Process

Key Metrics That Define "Large"

What LLMs Can Do

• Generate human-quality text and code
• Answer questions and explain concepts
• Translate between languages
• Summarize documents
• Reason through multi-step problems
• Follow complex instructions

Using an LLM via API

pythonCopy
from anthropic import Anthropic

client = Anthropic()

response = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Explain attention mechanisms in 3 sentences."}
    ]
)
print(response.content[0].text)

LLM Limitations

Popular LLMs (2025)