For clients who do not want to run model infrastructure themselves. Includes one managed model endpoint or assistant instance, monitoring, light updates, and fair-use support. Final model choice depends on hardware limits and client workload.
July 7, 2026