Green Fern
Green Fern

Adaptive Inference

AI Infrastructure

Large Language Models

Foundation models

AI models

LLMs

Generative AI

Open Source LLMs

Llama 4 Maverick and Llama 4 Scout now available on kluster.ai

Llama 4 Maverick and Llama 4 Scout now available on kluster.ai

By Anjin Stewart-Funai

Apr 9, 2025

Meta’s first open-weight natively multimodal LLMs usher in a new era of AI as they combine text and images in a single, efficient system. Now available on kluster.ai, Llama 4 Maverick and Llama 4 Scout give developers unparalleled performance for multimodal applications.

The Llama 4 revolution: Where Maverick and Scout fit in

Meta’s Llama 4 series introduces open-weight, multimodal models built on a mixture-of-experts (MoE) architecture. Among the first public releases in this family:

  • Llama 4 Maverick  (17B active parameters, 128 experts) 

  • Llama 4 Scout (17B active parameters, 16 experts)

Both models are distilled from Llama 4 Behemoth (288B parameters), Meta’s most powerful LLM to date, which has yet to be released. By leveraging its architecture, Scout and Maverick inherit Behemoth’s strengths in reasoning, coding, and multimodal tasks, while remaining far more efficient to deploy.

Llama 4 Maverick

Llama 4 Maverick is built for developers who need a high-performance, cost-effective multimodal model.

✅ Outperforms GPT-4o & Gemini 2.0 Flash in reasoning, coding, and multilingual tasks

✅ 400B total parameters (17B active per inference) for optimized efficiency

✅ Exceptional image understanding & creative writing capabilities

Llama 4 Maverick is ideal for powering advanced AI assistants and chatbots that require deep contextual understanding. It shines in multimodal use cases like text and image processing, as well as enterprise-grade AI solutions where top-tier reasoning and interpretability are essential.

Llama 4 Scout

Llama 4 Scout is purpose-built for ultra-long-context tasks, making it perfect for developers working with massive datasets, logs, or document streams.

✅ Outperforms Gemma 3, Mistral 3.1, and prior Llama models

✅ Strong image grounding and visual reasoning performance

Llama 4 Scout is a great fit for applications like multi-document summarization and retrieval, large codebase exploration, and building AI agents that need persistent memory. It also excels in enterprise search and knowledge management workflows that require detailed context and continuity.

Start building with Llama 4 on kluster.ai

Whether you’re creating advanced AI assistants, multimodal pipelines, or long-context enterprise tools, Llama 4 Maverick and Scout are now available on kluster.ai.