Runway Unveils New AI Video Model Gen 4.5 That Surpasses Google And OpenAI In Key Benchmark

Runway on Monday introduced Gen 4.5, a new text‑to‑video model that an independent benchmark ranks ahead of comparable offerings from Google and OpenAI. The company says Gen 4.5 generates high‑definition video from written prompts and demonstrates advanced comprehension of physics, human motion, camera movement, and causal relationships.

The model currently occupies the top position on the Video Arena leaderboard maintained by Artificial Analysis, an independent AI benchmarking firm. Rankings on the platform are determined by blind comparisons in which evaluators choose between two model outputs without knowing their origin. Google’s Veo 3 is listed second, while OpenAI’s Sora 2 Pro appears in seventh place.

Runway CEO Cristóbal Valenzuela told CNBC that the achievement underscores how a focused team can compete with much larger firms. Founded in 2018, Runway conducts AI research and develops video and world models trained on visual and observational data to better mirror real‑world dynamics. The company serves media organizations, studios, brands, designers, creatives and students, and PitchBook values it at $3.55 billion. Investors include General Atlantic, Baillie Gifford, Nvidia and Salesforce Ventures.

Valenzuela said the internal codename for Gen 4.5 was “David,” likening the effort to a David‑and‑Goliath story and describing the release as the result of years of work. He added that the company sees the current moment as one in which efficiency and focused research can broaden access to AI capabilities beyond a small number of dominant firms.

Runway plans a phased rollout of Gen 4.5, with availability for all customers by the end of the week. The model will be accessible via Runway’s platform, its API, and through selected partners, and it represents the first of several major releases the company has scheduled.