OpenAI released multiple API updates on Monday's Dev Day, officially launching its latest language model GPT-5 Pro, new video generation model Sora 2, and a more streamlined and cheaper speech model.
This series of initiatives is aimed at further attracting developers to join the OpenAI ecosystem, including the launch of the "AgentKit" tool for AI agent developers and new capabilities that allow the creation of applications within ChatGPT.
OpenAI CEO Sam Altman said that the arrival of GPT-5 Pro will attract developers who need "high accuracy and deep reasoning" in fields such as finance, law and medicine. Altman also emphasized that voice interaction is quickly becoming the main way for people to interact with AI, so OpenAI has launched a new version of the "gpt-realtime mini" voice model for API this time. This new model is not only smaller and lower in cost (only 30% of the price of the previous high-end voice model), but it still maintains "the same voice quality and expressiveness" and supports low-latency audio streaming interaction.
In addition, creators in the OpenAI developer ecosystem can now access Sora 2 through the API preview version. OpenAI last week released a new generation of audio and video generator Sora 2 and a supporting "TikTok-like" short video application Sora. Users can generate short videos of themselves or their friends based on prompt words and share them through dynamic streams recommended by the algorithm.
Altman said: "Developers can now use the same high-quality video generation model as Sora 2 in their applications." Compared with its predecessor, Sora 2 has more realistic and physically consistent scene performance, sound and picture synchronization, and significantly improved creative control - covering new capabilities from meticulous camera scheduling to stylized visuals. For example, users can expand the mobile phone's field of view to a "movie-level wide angle" and experience a dynamic audio-visual experience that is highly consistent with images, environmental sound effects, ambient sounds, and synchronization effects.
Sora 2 is positioned as a concept development tool that can be used for first draft creative ideas for advertising, and can also be used for applications such as toy designers turning sketches into product concepts. Altman pointed out as an example that this is a typical scenario where OpenAI joins hands with Barbie manufacturer Mattel to introduce generative AI into the behind-the-scenes process of toy design and content creation.