AI Model Compare / Llama

Llama: pricing, context window and best use

Llama by Meta. Current model: Llama 4 (Scout / Maverick). Best for: Open source / Self-hosting / Enterprise.

MakerMeta
Current modelLlama 4 (Scout / Maverick)
API input$0.08–0.27 / 1M tokens
API output$0.30–0.85 / 1M tokens
Context window1M–10M
Free tierFree open weights
Best forOpen source / Self-hosting / Enterprise
Official sitellama.com ↗

What Llama costs

API pricing is $0.08–0.27 per million input tokens and $0.30–0.85 per million output tokens. Subscription options: Free open weights to download · Pay-as-you-go via Groq / Together / Bedrock · Meta AI app free to use. These are list prices — caching, batch discounts and reseller contracts change the real bill. Figures were checked against official pages in September 2026.

Context window and limits

Llama has a 1M–10M context window. That ceiling covers everything in one request: system prompt, documents, retrieved passages and the conversation so far. Beyond it the input is truncated, so long jobs are usually split into chunks with a running summary.

What Llama is best at

Llama is built for Open source / Self-hosting / Enterprise. Because billing is per token, the effective cost depends on the input-to-output ratio: reading a lot and answering briefly is far cheaper than generating long output.

How Llama compares

Llama sits at the cheapest end of the price range covered here. The nearest alternatives by input price are Llama ($0.08–0.27/1M in) and DeepSeek ($0.15–0.66/1M in).

FAQ

How much does Llama cost per million tokens?

$0.08–0.27 for input and $0.30–0.85 for output, based on the list prices tracked on this page.

What is the context window of Llama?

1M–10M in a single request.

Is Llama free to use?

Free open weights. Meta publishes current terms at llama.com.

Other models on this page

Prices, context windows and plan limits are collected from each vendor's official pages and re-checked monthly. Last update: September 2026. This page is independent and not affiliated with Meta.