🦙

Meta Llama

Meta Llama is Meta's family of open-weight large language models, spanning Llama 3 through Llama 4 and beyond. It sits at #8 in real-world usage share with roughly 17.8%, and is arguably the single model most responsible for kickstarting the modern open-weight LLM ecosystem — thousands of fine-tunes, tools, and local-hosting setups (like Ollama) trace back to Llama. It's free to download and run on your own hardware, with no API key required.

Quick facts
Made by: Meta
Access: Open-weight, free to download and self-host (Ollama, vLLM, Hugging Face, llama.cpp)
Context window: ~128K tokens (Llama 3.x/4 class models)
Strengths: Strong general chat quality, wide ecosystem support, runs efficiently on consumer GPUs at smaller sizes
Primary use case: Self-hosted assistants, fine-tuning base for custom products, offline/private deployments

Jump to: Example Getting started Best for

Example

Run it locally with Ollama in one command, or call it through Ollama's local REST API from Python.

# run locally with Ollama
ollama run llama3.3

import requests

response = requests.post("http://localhost:11434/api/generate", json={
    "model": "llama3.3",
    "prompt": "Write a haiku about open-source software.",
    "stream": False
})

print(response.json()["response"])

Getting started

Install Ollama, pull a Llama model, and you're generating locally — no signup or API key needed.

# 1. install Ollama from ollama.com

# 2. pull and run the model
ollama pull llama3.3
ollama run llama3.3

Best for: Developers who want to fine-tune or fully own their model stack — self-hosted chat apps, on-device assistants, and products that need to run offline or keep data entirely in-house.