The groundbreaking natively multimodal Llama 4 has been officially unveiled, marking the dawn of a new era in the open-source community. At its inception, Llama 4 introduces two models: Scout and Maverick. Scout stands out as an industry pioneer, capable of supporting 10 million context runs on a single H100. Meanwhile, Maverick has emerged victorious over DeepSeek V3. Additionally, an ambitious supermodel boasting over 2 trillion parameters is still undergoing rigorous training. The arrival of Llama 4 has garnered immense attention and anticipation from all corners.
