Llama API pricing: Llama 4 Maverick, Scout and Llama 3.3 70B at every provider

Updated Oct 19, 2026

Llama models have open weights, so Meta is not the only seller: dozens of hosts and routers run them, and the price for the same model differs several times between them. The tables below are rebuilt from today's prices on every page view.

The models in short

In the 19 October 2026 collection, the lowest price per 1 million tokens, input / output:

• Llama 3.3 70B Instruct - from $0.08 / $0.26, 27 providers. Still the most offered Llama: a solid general model at a small-model price.
• Llama 4 Maverick 17B - from $0.15 / $0.20, 1 million tokens of context, 19 providers.
• Llama 4 Scout 17B - from $0.08 / $0.20, the longest context of the family (up to 1.3 million tokens), 17 providers.
• Llama 3.1 8B and Llama 3.2 1B / 3B - a few cents per million tokens: classification, extraction, routing.

Llama 3.3 70B: every provider

Live data: this table is built from today's prices.

Llama 4 Maverick: every provider

Live data: this table is built from today's prices.

Llama 4 Scout: every provider

Live data: this table is built from today's prices.

How to choose

• Lowest price, short prompts: Llama 3.1 8B or Llama 3.2 3B.
• General assistant at a low price: Llama 3.3 70B - compare the output price, it differs most between hosts.
• Long documents: Llama 4 Scout or Maverick, 1 million tokens and more.
• The cheapest row is not always the one to use: check the context the host actually serves, its quantization and the date on the card. All of Meta's models in one table: Meta Llama at every provider. To estimate a month, use the cost calculator.

All prices in US dollars per 1 million tokens, before taxes, as the providers publish them; every row links to a card with its source and check date.