Runway

A veteran AI video platform: its flagship Gen-4.5 tops the text-to-video leaderboard, with native audio and true-to-life physics—plus a real-time world model, GWM-1.

In the most cutthroat race in AI—video generation—Runway is one of the few old-guard players that has survived from 2018 to the present day and still sits at the center of the table. Gen-4.5, released in December 2025, put the company back on top: the flagship model claimed first place on Artificial Analysis’s text-to-video benchmark, generates 5-10 second HD clips with native audio, and its officially championed “trustworthy collisions” finally give object interactions a genuine sense of weight, rather than the usual clipping and ghosting artifacts.

What’s even more notable is that Runway’s ambitions now extend beyond “generating video.” Unveiled alongside Gen-4.5, the GWM-1 general world model supports real-time interaction at 720p and 24fps for minutes at a time. Users can steer the world being generated with camera poses, robot commands, or even audio, and its three variants target explorable environments, conversational video characters, and robotic manipulation simulation, respectively—this “video tools company” is now telling a story about embodied intelligence.

On the product side, Runway increasingly resembles a video generation aggregator: paid users can tap not only the in-house Gen-4.5, Act-Two motion capture, and Aleph video editing, but also third-party models such as Veo, Seedance, and Kling. Pricing starts at $12/month, and the free plan includes only a one-time grant of 125 credits—at the flagship model’s burn rate of 25 credits per second, a few trials will drain it dry.

For film and video workflow users and developers, Runway remains one of the most polished and most complete options available today—but the true cost of its credit system is worth calculating carefully before you subscribe.

Source: Agentopia · Full site machine catalog at /llms.txt