[10:42 Tue,2.December 2025 by Rudi Schmidts] |
Only two years after the release of Gen-1, Runway is now presenting version 4.5 of its well-known AI video generator, which is now facing serious competition from Google (Veo) and OpenAI (Sora). The infographic published by Runway comparing it to the competition as of November 30th shows a seemingly clear lead, but with a heavily scaled Y-axis: ![]() ELO Point Comparison AI Models as of November 30th In reality, the quality is likely comparable to other SOTA (State-Of-The-Art) models. This is nevertheless remarkable, as Runway has a relatively small development team with a significantly lower budget compared to Google or OpenAI. Anyone sifting through the news in the release blog will encounter many buzzwords that primarily demonstrate that there is relatively little new to announce. Ultimately, physics and object consistency are said to have been further improved:
![]() Runway Gen-4.5 More interesting are the limitations that the developers see in their own model: * Effects sometimes precede their causes (e.g., a door opens before the handle is turned). * Objects can unexpectedly disappear or appear in different frames (e.g., a cup vanishes after being obscured). * Actions are disproportionately often successful (e.g., a poorly aimed shot still scores a goal). However, these mentioned limitations are also found in the current competition. Nevertheless, the results of the new Runway Gen-4.5 model are quite impressive and are likely to cause a lot of eye-candy on the internet again. The first demo reel for the release naturally only shows cherry-picks—and thus the model at its best: Since the competition acts no differently with every new announcement, this still serves as a good comparison point for the current state of affairs—and at the same time, proof that Runway can continue to compete in the top league of AI models with the big players. deutsche Version dieser Seite: Runway Gen-4.5 ist da - wieder auf Augenhöhe mit Veo und Sora? |



