AI startup Runway launched Gen 4.5 on Monday, claiming the number one position on an independent video generation leaderboard and outperforming models from tech giants Alphabet and OpenAI. The New York-based company with just 100 employees now leads the Video Arena rankings maintained by Artificial Analysis, surpassing Google’s Veo 3 model in second place and OpenAI’s Sora 2 Pro in seventh.
The text-to-video model generates high-definition clips from written prompts and demonstrates what Runway describes as superior understanding of physics, human motion, and camera dynamics. Video Arena determines rankings through blind comparisons where users vote for their preferred output without knowing which company produced it, ensuring unbiased results.
David Versus Goliath in AI Video#
Runway CEO Cristóbal Valenzuela revealed his team deliberately codenamed the model “David” in reference to the biblical story. “We managed to out-compete trillion-dollar companies with a team of 100 people,” Valenzuela said. “You can get to new frontiers just by being extremely focused and diligent.”
The achievement marks a significant milestone for the startup, which secured $308 million in funding in April at a $3 billion valuation from investors including NVIDIA, SoftBank, Fidelity, and General Atlantic. Founded in 2018, Runway has now reached what PitchBook values at $3.55 billion.
Model Performance and Availability#
Gen 4.5 currently holds an Elo score of 1,247 on the Artificial Analysis benchmark. The model will roll out progressively to all Runway customers by the end of the week, with access available through the company’s platform, API, and partner integrations. Valenzuela indicated this release is the first of several major launches planned by the company.
Known Limitations#
Despite the advances, Runway acknowledges Gen 4.5 still struggles with object permanence and causal reasoning—common limitations across current video generation models. In some cases, effects can occur before their causes, a physics quirk that remains unsolved across the industry.
The ranking represents a rare win for a lean startup against the massive R&D budgets of Big Tech, demonstrating that focused execution can still compete at the frontier of AI development.