Runway Model Router
The Model Router picks the best model for your use case, on your rules for cost, quality and latency.
Get API KeyWhy use a Runway Model Router?
Get the right model, every call
Optimize for cost, quality or latency. Every call is matched to the model that best fits your priority and what you are creating.
You set the rules
Cap what a single generation can cost and exclude any model you would rather not use. The router only ever picks from what you allow.
See which model ran
Every response reports the model that served the request and what it cost. Automatic selection stays auditable.
Built on Years of Expertise
Runway has trained frontier models and served the best from other labs for years. The router encodes what we learned about which model wins which job, so the preference you set turns into a choice we have already tested.
New Models, Added Automatically
When a new model fits your rules, it becomes eligible automatically. Your routing keeps improving while your configuration stays the same.
Frequently
asked questions
Which models are available?
Video, image, audio and real-time models, including Gen-4.5, Seedance 2.5, Veo 3.1, GPT Image 2, Seedream 5 and ElevenLabs v3. The full catalog is above.
Do you support models from other labs?
The router can choose from supported Runway models and leading models from other labs. Your rules determine which are eligible.
When can I use a newly released model?
A new model can become eligible automatically when it fits your existing configuration, so routing improves without an integration change.
How much does a generation cost?
The cost depends on the model selected for the request. Every response reports the selected model and what the generation cost was.
How do I switch models?
Update your router configuration to change its priority, cost ceiling or eligible models. Your application keeps calling the same endpoint.
Can I try a model before integrating it?
Yes. Try supported models in the playground before integrating them into your application.
Which model should I use?
Set your priority for quality, speed or cost and the Model Router selects the eligible model that best fits each request.