Runway Gen-4.5 Beats OpenAI Sora 2 and Google Veo 3 in Text-to-Video Benchmark

Key facts

• Runway Gen-4.5 secured an Elo score of 1247, ahead of Google Veo 3 (1226) and OpenAI Sora 2 Pro (1206).
• The model delivers ‘unprecedented physical accuracy’ and visual precision, generating realistically moving objects and accurate dynamics for flowing liquid.
• Gen-4.5 handles styles from cinematic to photorealistic animation while maintaining the speed and efficiency of Gen-4.
• Built in collaboration with Nvidia, training and inference run on Blackwell and Hopper GPUs.
• The model is available across all subscription plans for creators and companies.