Build with real-time video, speech, and language models through shared APIs. Dedicated endpoints for growing workloads, plus model serving and distribution for model labs.