Choose between standard and priority throughput to match your workload's latency requirements.
Not yet available. Service tiers are on the roadmap.
Not every request has the same urgency. A batch job that runs overnight has different requirements from a real-time customer chat that must respond in under a second. Service tiers let you declare that priority explicitly and pay accordingly.
Standard is the default tier. Your requests join the general queue for that model and provider. Latency is good for most use cases — typically seconds for mid-sized completions — but during high-demand periods, requests may wait longer.
Standard tier is included with every plan at no surcharge.
Priority requests are routed ahead of the standard queue. They reach the model faster, which matters most for:
Priority tier carries a per-request surcharge on top of the normal token cost. The surcharge is shown in the model catalog for each model.
Pass "service_tier" in the request body:
| Value | Behavior |
|---|---|
"auto" | Standard tier (default) |
"standard" | Explicitly standard |
"priority" | Priority queue; surcharge applies |
The response body echoes the tier that was applied:
This is useful when using "auto" — you can confirm which tier was selected in logs.
Priority tier is only available on supported models. For models without a priority queue, "priority" silently degrades to standard. Check the service_tiers field in GET /api/models/{slug} to confirm support:
Priority surcharges are billed in DA alongside your normal token costs. The full breakdown — base token cost + priority surcharge — appears in your activity log per request.
There is no subscription required for priority access; you pay per use.