news
Runway stopped trying to win the leaderboard. Now it just resells whoever is winning it.
Every paid Runway plan bundles Veo 3.1, Kling 3.0 Pro and Seedance 2.0 alongside Runway's own models, all drawing from one credit pool. Its own Gen-4.5 fell out of the top 10 in seven months.
Runway's Gen-4.5 launched on 2 December 2025 at the top of the Artificial Analysis Text-to-Video leaderboard, scoring 1,247 Elo points ahead of Google's Veo 3 (1,226) and OpenAI's Sora 2 Pro (1,206). By May 2026, reporting placed it outside the top 10, displaced by Seedance 2.0, the then-anonymous HappyHorse-1.0, and the Kling 3.0 / Veo 3.1 cluster. By mid-July, it did not appear on the leaderboard at all.
The response was not a better model. It was a bigger menu.
Rather than chase the leaderboard with a Gen-5, Runway restructured its subscription around access rather than ownership. Every paid tier (Standard, Pro and Max) now includes Veo 3.1, Kling 3.0 Pro and Seedance 2.0 as third-party models, alongside Runway's own Gen-4.5, Gen-4 and Gen-4 Turbo, all drawing on a single shared credit balance. Reported per-second costs inside that pool run from 5 credits for Gen-4 Turbo up to 25 for Gen-4.5, with the third-party models priced separately within the same system.
The practical effect: a Runway subscriber no longer has to leave the product, or decide in advance which lab's model to trust, to access whichever model currently tops the rankings. Runway becomes the shelf, not the product on it.
Why this is the more important story than any single model launch
Our own platforms coverage has made the case that the model layer and the application layer in this market are two different businesses. Models compete on raw capability and get commoditised fast. The platforms that resell access to several of them at once are where the actual subscription revenue concentrates. Runway just made that argument for us, with its own roadmap: a company that used to be defined by owning a frontier model is now, by its own pricing page, mostly in the business of reselling its competitors'. Whether that is a stronger long-term position than owning the best model is exactly the kind of claim we would want measured, not assumed. It is one we plan to track as platform pricing and model access continue to shift.
Sources: Runway on X, VentureBeat.
Related: Runway
Priya Shenoy: Tracks what AI video actually costs across the platforms that resell access to the same handful of models. Treats a pricing page as a claim, not a fact, until someone checks it.