Build low-latency, multimodal experiences, including speech-to-speech. Build AI assistants within your own applications that can leverage models, tools, and knowledge to do complex, multi-step tasks. Run asynchronous workloads for 50% of the cost over 24 hours.