Meta released Llama 4 Scout and Maverick on April 5, 2026, the first open-weight models built on a mixture-of-experts architecture with text and vision tokens fused from pretraining. Scout packs 17B active parameters across 16 experts (109B total) and achieves an industry-leading 10M-token context window; Maverick uses 17B active parameters across 128 experts (400B total) and benchmarks above GPT-4o and Gemini 2.0 Flash on a wide range of tasks. Both models are available for download at llama.com and Hugging Face. A preview of Behemoth (~2T total parameters, 288B active) was also announced but not yet released.