Runway has launched its new Gen-4.5 AI video model, which has overtaken Google’s Veo 3 and OpenAI’s Sora 2 in a blind A/B testing leaderboard. This success highlights Runway's ability to compete against larger companies by focusing on video generation, particularly through metrics like prompt adherence and motion quality. The model, previously known as Whisper Thunder, emphasizes realistic physics and visual fidelity, marking a significant advancement in AI video technology. The development of Gen-4.5 relied heavily on NVIDIA's Blackwell architecture, optimizing inference to minimize costs associated with video generation. Runway's CEO Cristóbal Valenzuela pointed out that their concentrated effort allowed a small team to outperform massive corporations. This technical edge is evident in the model’s ability to simulate realistic object dynamics and fluid movements, which are often challenging for generative video technologies. On the same day as Runway's launch, ByteDance and Tencent introduced their open-source models, Vidi2 and HunyuanVideo-1.5, respectively. Both models target different market segments: ByteDance emphasizes editing and understanding, while Tencent aims to make high-quality video generation accessible on consumer hardware. This creates a clear divide in the market, with Runway focusing on premium, closed solutions and its competitors pushing democratized access to AI tools. Runway's independence from Meta following a failed acquisition bid appears to have been a strategic move. By not selling out, Runway is building a cohesive ecosystem that integrates Gen-4.5 with its Aleph video editor, offering a streamlined workflow for users. The pricing for Gen-4.5 remains competitive, aiming to maintain market share against the growing presence of open-source alternatives.