Llama 4 Maverick is a multimodal LLM with 17 billion active parameters, built on a mixture-of-experts (MoE) architecture with 128 experts and a total of 400 billion parameters. It is designed for balanced, high-quality performance across text, image understanding, reasoning, coding, and multilingual tasks.

It excels in delivering strong results on chat, code generation, complex reasoning problems, long-context retention (up to one million tokens), and advanced image analysis—outperforming or matching leading models like GPT-4o and Gemini 2.0 Flash in many benchmarks. Its MoE design enables efficient inference by activating only the most relevant subset of experts per token.

Some other noteworthy features of Llama 4 Maverick include native support for both text and images as input (enabling document intelligence use cases), robust multilingual capabilities across at least a dozen languages, and suitability for enterprise-scale applications such as customer support bots that process screenshots or creative assistants that generate content from multimodal documents.

Metric	Value
Parameter Count	400 billion total / 17 billion active
Mixture of Experts	Yes
Active Parameter Count	17 billion
Context Length	1 million tokens
Multilingual	Yes
Quantized*	No

*Quantization is specific to the inference provider and the model may be offered with different quantization levels by other providers.