Runway Gen-4.5 API pricing

Configuration-specific buying guide

How to use this Runway Gen-4.5 API pricing result

This page answers one narrow procurement question: what do documented API channels charge for an output second from the exact Runway Gen-4.5 model? The owner route lists model key gen4.5 at twelve developer credits per generated second, and the developer credit price is one cent. Replicate lists its exact runwayml/gen-4.5 route at the same twelve-cent output-second rate. These are comparable purchase channels, but they are only two observations, so APIDir displays both and deliberately leaves the ranking unpublished.

A Runway web subscription is not an API quote. Subscription bundles, Explore Mode, promotional credits, and the developer portal use different purchasing rules. Gen-4 Turbo is also a different model even though its name is close and its per-second rate is lower. Folding any of those numbers into this page would answer a different question and manufacture a misleading bargain.

Price parity does not make the channels operationally identical. Account setup, request schema, concurrency, queue behavior, retention, moderation, support, and error semantics can differ. Treat the table as a verified meter result, then benchmark the route that fits your deployment and review the source again before committing budget.

What changes the real cost

• Generated duration is the primary billable quantity; budget accepted clips, deliberate reruns, and quality-control retries.
• Text-to-video and image-to-video can share a model price while requiring different inputs, validation, and storage work.
• Developer API credits must not be replaced with an effective web-subscription credit price or an unlimited-plan assumption.
• Egress, retained assets, moderation failures, post-processing, and human review remain outside the displayed output-second meter.

Checks before production

• Pin the exact gen4.5 or runwayml/gen-4.5 route instead of a Runway family alias.
• Validate duration and aspect ratio before submission, and reject unsupported values before a billable request is created.
• Use idempotency for job creation where supported and bounded backoff for status reads; never retry an unknown submission blindly.
• Record route, output seconds, provider request ID, terminal state, and charged units so invoices can be reconciled.

Workloads this result can inform

Ten-second campaign clips

Multiply the verified per-second meter by ten, then include the expected rejection and rerun rate. The table alone describes a successful output, not the full creative funnel.

Provider failover

A second channel can reduce account or availability concentration, but test whether request fields, moderation responses, and output handling remain compatible before enabling automatic fallback.

Subscription versus API planning

Keep creator-seat economics and programmatic developer usage in separate budgets. A web plan may suit manual ideation while an API route serves an automated product.