Runway Gen-4.5 is rolling out, and the text-to-video model grabbed the No. 1 spot on the Video Arena leaderboard for generative models that don’t simultaneously output sound, beating the non-audio versions of Google’s Veo 3 and OpenAI’s Sora 2 Pro. Offering what Runway AI calls “unprecedented visual fidelity” across cinematic and “highly realistic” outputs, it is also built for creative freedom, providing “precise control over every aspect of generation.” The new model is good at understanding physics, cause and effect through casual reasoning, camera movements and human emotion, claims the New York-based startup.

Runway plans to make Gen-4.5 “available at comparable pricing across all subscription plans” and will be available through the company’s web platform as well as its API.

The rollout is staggered, with the new product expected to be available to all users by the end of the week. And though it’s starting with written prompts, the company plans to eventually bring all existing control modes (image-to-video, keyframes, video-to-video and more) to Gen-4.5.

“We managed to out-compete trillion-dollar companies with a team of 100 people,” Runway CEO Cristóbal Valenzuela told CNBC, adding that users “can get to frontiers just by being extremely focused and diligent.” The executive calls Gen-4.5 “an overnight success that took like seven years.”

According to a Runway research post: “Gen-4.5 achieves unprecedented physical accuracy and visual precision. Objects move with realistic weight, momentum and force. Liquids flow with proper dynamics. Surface details render at great fidelity. And fine details like hair strands and material weave remain coherent across motion and time.”

However, The Verge points out “some limitations,” including possible “issues with object permanence and causal reasoning, meaning effects may happen before the cause, such as a door opening before someone uses a handle.”

The Verge also cites Runway’s “claims that photorealistic visuals created with Gen-4.5 can be ‘indistinguishable from real-world footage with lifelike detail and accuracy.’” The company says it can deliver stylized images that mimic anime and paintings or illustration.

“It does feel like a very interesting moment in time where the era of efficiency and research is upon us,” said Valenzuela. His company is “excited to be able to make sure that AI is not monopolized by two or three companies,” as evidenced by the Video Arena chart published by Artificial Analysis.