LiveKit Inference is a newly launched low-latency model gateway specifically designed for voice AI, allowing seamless access to top-performing speech-to-text, large language models, and text-to-speech models using just a LiveKit API key. This platform eliminates the need for managing multiple provider accounts and integrations by partnering with leading model providers like OpenAI and AssemblyAI, offering a consistent API for simplified model access and billing. It addresses common challenges such as concurrency management and latency optimization, providing a unified dashboard for visibility into usage limits and leveraging global co-location and provisioned capacity to reduce latency and enhance reliability. LiveKit Inference is immediately accessible through both Python and Node Agents SDKs, aiming to streamline the development of voice-driven AI products by handling the infrastructure complexities and enabling developers to focus on innovation.