GPT4o (Omni)

GPT-4o (Omni) is OpenAI's first end-to-end multimodal model that processes text, vision, and audio within a single neural network. Trained across all three modalities, it enables real-time, low-latency voice conversations, visual understanding of images and on-screen content, and expressive speech synthesis. The model supports natural turn-taking, handles interruptions, and is robust to accents and background noise. GPT-4o is also more cost-effective, priced at half the cost of GPT-4 Turbo, and can generate 3D images as shown in examples. While the API initially supports text and image inputs, the full audio and vision capabilities are expected to follow. GPT-4o is designed for developers, customer support, education, accessibility, and any scenario requiring fluid human–AI interaction across speech, text, and visuals.