Llama 3.1, a collaboration between Fireworks and Meta, introduces advanced AI capabilities with features such as expanded context length, multilingual support, and tool-calling, making it a significant advancement in AI technology. Fireworks, as a launch partner, offers immediate access to Llama 3.1 through its serverless inference engine, optimized for low latency and efficient deployment. The platform supports the creation of compound AI systems by integrating multiple models and tools, enhancing performance, reliability, and control. Llama 3.1, with its unmatched flexibility and control, supports context lengths up to 128K across eight languages, enabling innovative applications in synthetic data generation and model distillation. Fireworks emphasizes a commitment to open and responsible AI development, providing robust tools for developers to create custom AI applications while ensuring security and safety measures. Additionally, the introduction of AMD Instinct MI300 accelerators alongside NVIDIA H100 further powers the serverless inference capabilities for Llama 3.1, positioning Fireworks at the forefront of AI innovation.