Planned for Dec 2026
Semantic routing
Route each request to the lowest-cost model that answers it well.
For mixed workloads.
2026–2027 delivery plan
Inference and Nexus Composer are available in pilot today. Training is in private preview. GPU capacity and agent tooling are next.
Planned releases
The next services extend the platform pilot teams use today.
Planned dates may change.
Route each request to the lowest-cost model that answers it well.
For mixed workloads.
Run variable AI workloads without maintaining a fleet.
For event-driven demand.
Test tools, workflows, and model behaviour in a contained environment.
For agent teams.
Reserve capacity for sustained workloads that need predictable access.
For committed capacity.
Stay informed
Product milestones, preview invitations, and availability updates only.