Runway launches Media Router to pick your AI model so you don't have to

Runway doesn’t want you to care which AI model generates your video. It just wants it to work. On Thursday, the startup launched Runway Media Router, a tool that automatically picks the best image, video, or audio model for a given task based on whether a developer prioritizes quality, speed, or cost. It’s an admission that the generative media landscape has gotten too fragmented for most teams to track — and a bet that infrastructure, not just model performance, is where the money is.

The router lives inside Runway Dev, the company’s newly launched developer platform that already counts Adobe, Cloudflare, ElevenLabs, Expedia, Shutterstock, and Quora as customers. These companies can now integrate media generation directly into their own products via API rather than pointing users to Runway’s app. Anthony Maggio, Runway’s chief product officer, framed it as a one-stop shop: “Most developers are not spending the time to really understand the capabilities of each of these models and where they excel or differ.” The router’s intelligence layer is built on what Runway’s in-house creative team has learned evaluating output across media types — how a video model handles motion, how an image model handles composition, how a voice model handles lip syncing.

The launch also nods to geopolitics. Maggio noted that developers can set preferences for American model providers, a filter that may grow in demand as the Trump administration explores bans on Chinese open AI models. But the more immediate pressure is price. Token pricing became a flashpoint in 2026 after enterprises got burned by ballooning agentic AI bills. Runway itself just swapped unlimited subscriptions for token-based pricing, a move that drew user backlash. Routing for cost, Maggio says, is what customers actually care about.

This pivot also reflects a competitive reality. Runway hasn’t released a new frontier video model since Gen 4.5 in December. While its Aleph 2.0 video editing model still ranks near the top, its text-to-video and image-to-video models have been overtaken by Google, ByteDance, and Alibaba. Runway’s answer is to stop asking developers to bet on a single winning model and instead sell them the switchboard. The same routing tech that powers Runway’s own AI agent product is now being packaged for outsiders. I’m watching whether developers actually trust a third party to make their model choices — that’s a lot of abstraction to hand over.