The End of Sora - OpenAI Makes Drastic Priority Shift
Ronni Holmvig Strøm · 2026-03-25
OpenAI has quietly begun winding down its standalone AI video generation product, Sora.
OpenAI has quietly begun winding down its standalone AI video generation product, Sora.
The move, confirmed internally this week and first reported by The Information on March 18, marks the end of Sora as an independent consumer-facing tool.
Instead, OpenAI is redirecting the underlying video-generation technology into its core products, particularly GPT-5.4’s computer-use capabilities and future agentic workflows.
The decision comes less than 14 months after Sora’s dramatic debut in February 2025, when its photorealistic, physics-aware video clips captured global attention.
What started as one of the most hyped consumer AI launches of the year is now being deprioritized in favor of agent-focused development.
Why Sora Is Being Sunsettled
According to sources familiar with the decision, OpenAI’s leadership concluded that standalone video generation, while technically impressive, sits outside the company’s highest-ROI priorities for 2026. The strategic focus has shifted decisively toward:
Long-horizon reasoning and autonomous agents.
Native computer-use and desktop automation (expanded in GPT-5.4).
Enterprise-grade tool integration and persistent workflows.
Video generation is not being abandoned entirely. The Sora model family and its underlying world-model research will continue, but as supporting components inside agentic systems rather than as a standalone product. For example, future agents may use Sora-derived capabilities to generate short explanatory clips, simulate physical actions, or create training data for robotics, but users will no longer access a dedicated “Sora” interface for general video creation.