SAN FRANCISCO — OpenAI introduced a slate of major updates at its Dev Day event Monday, unveiling GPT-5 Pro, its most advanced language model to date, along with a new video generation model, Sora 2, and a smaller, cheaper voice model designed for real-time interactions.
The announcements were part of a broader effort to attract developers to OpenAI’s growing ecosystem. The company also debuted a new agent-building tool and features that allow developers to build applications directly within ChatGPT.
GPT-5 Pro is aimed at industries that require high accuracy and complex reasoning, such as finance, healthcare, and law, according to OpenAI CEO Sam Altman.
Altman emphasized the growing importance of voice capabilities in human-AI interaction, announcing “gpt-realtime mini,” a smaller and more affordable voice model that supports low-latency streaming for speech and audio. The new model is 70% cheaper than OpenAI’s previous advanced version but maintains similar voice quality and expressiveness.
Developers can also access Sora 2, OpenAI’s next-generation video and audio generator, now available in preview through the API. Sora 2 builds on its predecessor with more realistic, physically consistent scenes, synchronized sound, and enhanced creative control — including detailed camera directions and stylized visuals.
“For example, you can take the iPhone view and prompt Sora to expand it into a sweeping, cinematic wide shot,” Altman said. “One of the most exciting improvements is how well this model pairs sound with visuals — not just speech, but rich ambient audio and synchronized effects grounded in what you’re seeing.”
Sora 2 is positioned as a tool for concept development, helping creators generate visual prototypes, ad storyboards, or product concepts. During his presentation, Altman highlighted a collaboration with Mattel that uses Sora to turn sketches into toy designs, illustrating how generative AI is beginning to reshape creative industries.