At its Dev Day event on Monday, OpenAI introduced several API enhancements, unveiling GPT-5 Pro—their newest language model—alongside Sora 2, a next-generation video creation model, and a more compact, cost-effective voice model.
These updates are part of a broader effort to attract developers to OpenAI’s platform, which also included the debut of a tool for building agents and the capability to develop applications directly within ChatGPT.
According to OpenAI CEO Sam Altman, GPT-5 Pro is likely to attract developers working in sectors like finance, law, and healthcare—fields that demand a high level of reasoning and precision.
Altman further emphasized that voice interaction is rapidly becoming a key method for engaging with AI. In response, OpenAI is introducing “gpt-realtime mini,” a streamlined and more affordable voice model available via API, designed for fast, low-latency audio and speech streaming. This new model is 70% less expensive than OpenAI’s earlier advanced voice model, yet still delivers equivalent voice quality and expressiveness.
Additionally, developers within OpenAI’s ecosystem can now preview Sora 2 through the API. Sora 2, the latest audio and video generation tool, was launched last week together with the Sora app—a TikTok-like platform featuring short, AI-generated videos. The Sora app enables users to create videos of themselves, friends, or anything else from a prompt, sharing them through a feed similar to TikTok’s algorithmic stream.
“Developers now have the opportunity to use the same model that drives Sora 2’s impressive video results directly in their own applications,” Altman stated.
Building on its predecessor, Sora 2 delivers more lifelike and physically accurate scenes, complete with synchronized audio and enhanced creative options—from intricate camera movements to unique visual styles.
“For instance, you can prompt Sora to transform an iPhone shot into a dramatic, cinematic wide angle,” Altman explained. “But what excites us most is how this new model aligns sound with visuals—not just dialogue, but immersive soundscapes, ambient noises, and effects that match what’s happening on screen.”
Sora 2 is positioned as a tool for conceptual development, whether it’s generating a visual concept for an advertisement based on a product’s mood, or enabling a Mattel designer to turn a sketch into a toy prototype—an example Altman shared at Dev Day, highlighting OpenAI’s collaboration with the Barbie-maker to integrate generative AI into toy design workflows.
