Runway model router cost: billed at the model it picks
Runway bills a router generation at the rate of the model the router selects and reports model and cost. Sume's sume/auto echoes its id and reports usage.cost.

A Runway Model Router generation costs the standard rate of whichever model the router selects, and the response metadata reports the model used and the realized cost in credits. Sume's sume/auto works differently on one point: responses echo sume/auto and never disclose which family ran, but usage.cost still gives the billable amount.
How does Runway bill a router generation?
The Runway Dev pricing guide says generations routed through a Model Router "are billed at the standard rate of whichever model the router selects." Credits are bought for $0.01 each, so the cost of a routed job is the selected model's credits times one cent. The guide gives the metadata promise in one sentence: the response reports the model used and the realized cost in credits. Read 2026-10-01.
What does Sume report for sume/auto?
In Sume's video generation docs, model: "sume/auto" lets Sume pick the family. The table row reads: "Responses echo sume/auto; the family that ran is never disclosed." Billing is in USD: the workspace balance is reserved on submit at provider list × 1.25, and usage.cost is the Sume billable amount. So you can audit what a job cost without learning which model produced it.
| Question | Runway Model Router | Sume `sume/auto` |
|---|---|---|
| Rate applied | Standard rate of the selected model | Provider list × 1.25, reserved on submit |
| Unit | Credits ($0.01 each) | USD from the workspace balance |
| Model named in response | Yes, model used is reported | No, sume/auto is echoed |
| Cost named in response | Realized cost in credits | usage.cost |
Can I see the cost before a job runs?
Through MCP, yes. Paid tools accept dry_run=true, which the tools and gates docs describe as "Admission/cost preview only; do not submit the job." An optional max_spend_usd is enforced when you provide it. If you need to know which model ran, name the model in the request instead of routing; see Runway API pricing per second for the vendor side of per-model rates.
Which approach fits which need?
If you need a per-job record of which model was used, Runway's metadata gives that. If you want one id that picks for you and a single cost field to reconcile, sume/auto with usage.cost does that. Pin a model id when the choice matters to the output.
Sources
Related posts
More in Pricing
- Runway multi_shot_video at 13 credits per second vs Sume clips
Runway prices multi_shot_video at 13 credits per second at 720p, 17 at 1080p. On Sume you generate clips and join them with Timeline; costs for 15 seconds.
- Runway ProRes and PNG sequence: 5 credits per second extra
Runway adds 5 credits per second for ProRes and png_sequence, 20 for 10-bit or HDR, on Gen-4.5 and Aleph 2.0. A 10-second example and Sume's video pricing.
- Runway Seedance 2 at 40 credits per second vs Sume at 1080p
Runway lists seedance2 at 36 credits per second for 480p and 720p, 40 at 1080p and 150 at 4K. Sume's seedance-2 reserves list x 1.25 and tops out at 1080p.
- Runway Seedance 2 at 36 credits a second vs Sume's Seedance rates
Runway lists seedance2 at 36 credits per second and seedance2_fast at 29 for 480p and 720p. Sume's seedance-2 and seedance-2-fast price by resolution.
Written by Sume