Llama is made by Meta. The weights are open, so you can run it yourself, but that takes a strong server. On NeuralBox it runs in the browser.
Llama 4 Scout holds more than a million tokens of context. Maverick gives better answers. Llama 3.3 70B fits simple tasks.
Available versions
Llama 4 Maverick
per exchange ≈ $0.002
The main Llama 4 model. Context 1M tokens.
Llama 4 Scout
per exchange ≈ $0.001
The longest context. Context 1.3M tokens.
Llama 3.3 70B Instruct
per exchange ≈ $0.001
The previous generation. Context 131K tokens.
What it does
Cheap
A fraction of a cent per full answer.
Long context
Scout holds over 1M tokens.
Images
Llama 4 understands images.
Prompt examples
Attach files and images to any message.
Email
Write a polite no to a supplier and offer a call next week
Translation
Translate this text into formal business Spanish
Parameters
Billing
per answer, no subscription
Attachments
files and images in a message
Switching
change the model any time
Context
up to 1.3M tokens (Scout)
How much Llama costs
Model
Exchange, tokens
USD
Llama 4 Maverick
64
≈ $0.002
Llama 4 Scout
30
≈ $0.001
Llama 3.3 70B Instruct
32
≈ $0.001
One exchange here means a 1,000-token question and a 1,000-token answer, about 700 words together. USD at the Basic rate: $5 = 150,000 tokens. Long threads and model reasoning cost more. The exact price of each answer shows in the chat.
$5 gives 150,000 tokens. With Llama 4 Scout that is about 5,000 questions with a full answer.
How to start
Sign in
With Google, Telegram or email. Takes under a minute.
Top up
From $5 by card or from $2 in crypto. No subscription, and tokens never expire.
Generate
Open the Text tab and pick Llama. Ask a question or attach a file.