Meta's open-weight mixture-of-experts family with native multimodality and very long context.
Specifications
- Provider
- Meta AI
- Type
- Open-source / open-weight
- Modality
- Multimodal (MoE)
- Category
- Multimodal model
- Context window
- Up to 10M
- License
- Llama Community
- Released
- April 5, 2025
What it was trained for
Meta Llama 4 is an open-weight, natively multimodal model family using a mixture-of-experts architecture, trained to handle text and image inputs with very long context windows.
Best for
- ▸Long-context document and conversation processing
- ▸Multimodal text-and-image applications
- ▸Self-hosted and customizable deployments
- ▸Fine-tuning for domain-specific assistants
- ▸Multilingual generation and understanding
Capabilities
Open weightsMixture-of-experts architectureNative multimodal inputVery long context windowMultilingual support
Performance & positioning
A capable open-weight family whose mixture-of-experts design provides strong quality with efficient active-parameter usage at inference.
More from Meta AI