Streaming in AI on Demand

We have enabled optional streaming for chat completions in AI on Demand. Responses can now be delivered progressively rather than as one complete block at the end.

The main benefit is perceived responsiveness. Streaming does not make the model generate faster, but it allows applications to display text as it becomes available. This is particularly useful for chat interfaces, copilots and content assistants, where users can begin reading without waiting for the full answer.

Streaming uses the existing API endpoint and authentication. Existing integrations continue to work unchanged, and complete responses remain available wherever they are preferred.

Further information on using streaming, as well as practical examples for the console, can be found in our documentation