Runway is giving its AI video lineup a quiet flex with Gen-4.5, a model that recently topped similar offerings from Google and OpenAI in an independent benchmark.
The platform lets users turn written prompts into high-definition video, showing a stronger understanding of physics, human motion, camera movement, and cause-and-effect than earlier iterations.
For creative teams and enterprise users, the update aims to make AI-generated footage feel more intentional and less like rough drafts. Gen-4.5 helps Runway keep its work at the forefront as the competition for AI video generators heats up.
Visual realism and prompt fidelity
In its official announcement, Runway said Gen-4.5 produces “cinematic and highly realistic outputs,” claiming improvements across physics, motion, and visual consistency. The latest version also maintains the speed and efficiency of Gen-4, delivering breakthrough quality without compromising performance.
Runway highlighted that one of the new model’s standout features is its prompt fidelity, which produces physically accurate and visually precise AI-generated footage. As Runway emphasized, “objects move with realistic weight, momentum, and force.” The company added that the advancement extends to fluid dynamics, surface rendering, and fine details like hair and fabric textures, which now maintain coherence across motion and time.
According to CNBC, Runway’s new model ranks at the top of the Video Arena leaderboard, an independent ranking maintained by AI benchmarking firm Artificial Analysis. The leaderboard is based on a blind comparison system where evaluators choose between two video outputs without knowing which companies produced them.
Additionally, with 1,247 Elo points, Gen-4.5 currently holds the top spot in the Artificial Analysis Text to Video benchmark, surpassing all other models.
Benchmark standing and competitive position
Runway framed Gen-4.5 as a meaningful milestone for its long-term research efforts. CNBC also added that Runway’s lean team structure helped it iterate faster on specialized model development, a contrast to the broader multimodal focus seen at larger AI labs.
CEO Cristóbal Valenzuela told CNBC, “We managed to out-compete trillion-dollar companies with a team of 100 people.” Valenzuela also added that teams can reach the frontiers just by being extremely focused and diligent, saying that the model was internally codenamed “David,” a reference to the David and Goliath story.
Runway views Gen-4.5 as part of a broader push to expand its video and world-model research, with more releases planned.
Customer base, rollout, and what comes next
CNBC highlighted that Runway’s customer base includes media organizations, studios, brands, designers, creatives, and students, and that its valuation rose to $3.55 billion, citing PitchBook. The company, founded in 2018, also earned a place on CNBC’s Disruptor 50 list and has investors including General Atlantic, Baillie Gifford, Nvidia, and Salesforce Ventures.
Valenzuela told CNBC that Gen-4.5 would roll out gradually and would be available to all customers by the end of the week. He added that it will be accessible through Runway’s platform, its API, and partner integrations. The publication also noted that Gen-4.5 is the first of several major releases as the company works to ensure AI development is not “monopolized by two or three companies.”
Runway also shared technical details about Gen-4.5’s performance infrastructure, saying that the model was developed entirely on Nvidia GPUs across research, pre-training, post-training, and inference.
The company said that it collaborated extensively with Nvidia to improve training efficiency and inference speed in its video diffusion system. “Inference for Gen-4.5 runs on Nvidia Hopper and Blackwell series GPUs, enabling optimized performance without sacrificing output quality,” said Runway.
As more advanced systems like Gen-4.5 reach commercial platforms, the broader market for AI-generated art and video appears to be moving beyond early free experimental tools toward more capable paid systems that prioritize reliability, control, and production-ready output.
For a broader look at how Runway’s previous model performs in practice, read our full review.