Llama 4 Scout
Neural Network
Llama 4 Scout API is Meta's multimodal mixture-of-experts model for text tasks and image processing.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 1 310 720 tokens
- Max answer length
- 16 384 tokens
- Capabilities
- function calling
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Llama 4 Scout?
Llama 4 Scout is a Meta language model based on a mixture-of-experts architecture: out of 109 billion parameters, 17 billion are activated per request, distributed across 16 experts, resulting in a lower load than a dense model of the same size. It is multimodal, accepting text and images, and responding with text. In BotHub, it is accessible from Russia without a VPN or foreign card; payment is made with Russian cards in rubles, charged per token. Over 250 neural networks are available in one window, allowing you to switch between them directly in the chat. A unified OpenAI-compatible API lets you change models in your service without rewriting the integration. Chats are encrypted, data is not saved, and companies have access to contracts, invoices, electronic document management, and an admin panel with limits. It is useful for analyzing and summarizing long documents, setting up a chat assistant for repetitive queries, extracting data from screenshots, diagrams, and photos, drafting emails or descriptions, and building multimodal products where some requests go to other models.Frequently asked questions about Llama 4 Scout
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
It is a text model with a 131,072-token context: it answers questions, writes and edits code, analyzes documents and images, and helps with long-form material analysis. Suitable for chats and support, processing large files, preparing drafts, and embedding into services via the unified BotHub API.
The model generates up to 16,384 tokens per response—enough for a lengthy article, detailed instruction, or large code snippet. If you need longer text, continue the dialogue in parts: the 131,072-token context allows you to keep the entire material in front of the model.
A separate reasoning mode is not noted in our data, so do not expect hidden chains of thought by default. However, you can explicitly ask it to break down a task step-by-step in your prompt—this noticeably improves accuracy for logic, calculations, and code analysis.
Yes, function calling and structured JSON output are supported. The model returns answers according to a specified schema and accesses your tools: search, databases, calculators, and internal services. This is key for agents and automation, and the OpenAI-compatible BotHub API connects without rewriting integrations.
Images and documents—yes: send a screenshot, photo, diagram, or file and ask it to describe the content, extract data, or compare versions. Audio and video input is not noted in our data, so for sound and clips, choose a specialized model from the BotHub catalog.
There are no notes on specific language support in our data, so we will not evaluate this. It is more reliable to test it in practice: send your typical query and compare the response with another model in the same window; switching takes seconds, and you only pay for the tokens used.