OpenAI ramps up developer push with more powerful models in its API (techcrunch.com)

At its Dev Day, OpenAI pushed more powerful tools into its API: GPT-5 Pro, the new multimodal video generator Sora 2 (now available in preview), and a compact voice model called gpt-realtime mini. GPT-5 Pro is positioned for applications that demand high accuracy and deep reasoning—think finance, legal, and healthcare—while gpt-realtime mini enables low-latency streaming audio/speech interactions at about 70% lower cost than OpenAI’s prior advanced voice model, claiming the same voice quality and expressiveness for real-time voice-first experiences. Sora 2 brings more realistic, physically consistent scenes with synchronized sound and fine-grained creative controls (detailed camera direction, stylized visuals, ambient audio and effects). Developers can now use the exact model behind OpenAI’s Sora app to generate short, shareable videos—making it easier to prototype ads, concept visuals, or product design assets (OpenAI highlighted toy-design workflows with partners). Taken together, these API updates lower the barriers to integrating advanced multimodal generation and real-time voice in production apps, accelerating the shift to voice-and-visual-first interfaces and raising the competitive stakes for other AI/ML platform providers.