Llama 4 Maverick

Llama 4 Maverick, released by Meta on April 5, 2025, was among the first open-weight models to combine natively multimodal understanding with a Mixture-of-Experts architecture — accepting text, images, and other inputs without separate specialized modules. Its release through Meta's open-weight program made highly capable multimodal MoE architecture accessible to researchers and organizations deploying models on their own infrastructure, continuing Meta's strategy of advancing the open-weight ecosystem with frontier-class architecture.