llamafile

One file that is both the model and the runtime. Download, chmod +x, run. No install at all.

llamafile is Mozilla’s trick: one file that is both the GGUF weights and the runtime (llama.cpp plus Cosmopolitan Libc). Download, chmod +x, run. No install, no Python, no Docker.

`--server` exposes OpenAI-compat on port 8080. Context is -c at launch. The right tool when you want to hand someone a model as an executable.

Max context: Set with -c at launch
Free tier: Unlimited and private
Requirement: Nothing — no account

Endpoint

Base URLhttp://localhost:8080/v1

Rate limitNone — it is your machine

OpenAI-compatible
text

Code examples

npm install openai

import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'http://localhost:8080/v1',
  apiKey: 'not-needed'
})

const response = await client.chat.completions.create({
  model: 'LLaMA_CPP',
  messages: [{ role: 'user', content: 'Explain closures in one paragraph.' }]
})

console.log(response.choices[0].message.content)

pip install openai

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8080/v1",
    api_key="not-needed",
)

response = client.chat.completions.create(
    model="LLaMA_CPP",
    messages=[{"role": "user", "content": "Explain closures in one paragraph."}],
)

print(response.choices[0].message.content)

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "LLaMA_CPP",
    "messages": [{"role": "user", "content": "Explain closures in one paragraph."}]
  }'

Worth knowing

`chmod +x model.llamafile && ./model.llamafile --server` and you are done.

Models you get for free2

Llama 3.3 70Bllama-3.3-70b
Mixtral 8x7Bopen-mixtral-8x7b