OpenAI courts developers with GPT-5 Pro, Sora 2 and low-latency voice AI

October 7, 2025

editorial_staff

OpenAI used its Dev Day to roll out a slate of API updates designed to attract more developers to its platform. The company introduced GPT-5 Pro, the latest flagship language model; Sora 2, a new audio-and-video generator available in preview through the API; and “gpt-realtime mini,” a smaller, lower-cost voice model. OpenAI also announced an agent-building tool and the ability to build apps inside ChatGPT, underscoring a push to make its ecosystem the default place to create AI products.

GPT-5 Pro is positioned for teams that need high accuracy and stronger reasoning, with OpenAI highlighting use cases in finance, legal services, and healthcare. CEO Sam Altman said voice will be central to how people work with AI, and the new gpt-realtime mini aims to support that shift. The model offers low-latency streaming for audio and speech, comes in at 70% less than the company’s previous advanced voice model, and is billed as delivering the same voice quality and expressiveness.

Sora 2 extends OpenAI’s media tools. Developers can now tap the same model that powers Sora 2’s video outputs within their own apps, while the consumer-facing Sora app, launched last week, lets users create short AI videos and share them in a TikTok-style feed. Sora 2 promises more realistic, physically consistent scenes with synchronized sound and gives creators finer control, from detailed camera directions to stylized looks. Altman noted that prompts can expand a simple iPhone frame into a cinematic wide shot, and emphasized how the model pairs visuals with rich soundscapes, ambient audio, and synchronized effects grounded in what appears on screen.

OpenAI is pitching Sora 2 as a tool for concept development across industries, from early ad visuals to product design. Altman pointed to a Mattel example, where a designer could turn a sketch into a toy concept, illustrating how generative AI could move into creative pipelines. Together, the releases signal a broader bid to equip developers with more capable text, voice, and video building blocks inside OpenAI’s API.