OpenAI updates tools for voice, agents, and browser-based work
OpenAI announced new Realtime models, a Chrome extension for Codex, Agents SDK updates, GPT-5.5, and GPT-Image-2.
Summary
OpenAI sent a developer update with several launches focused on real-time voice, agents, and multimodal creation.
Highlights include new Realtime API models such as gpt-realtime-2, gpt-realtime-translate, and gpt-realtime-whisper, a Chrome extension for Codex, updates to the Agents SDK, GPT-5.5, and GPT-Image-2.
In practice
The new Realtime models are designed to support more natural voice agents, with stronger instruction following, multilingual capabilities, live translation, and word-by-word transcription as speech happens.
The Codex Chrome extension allows the agent to work directly inside websites where the user is signed in, organize task-specific tabs, and return results for review. OpenAI also highlighted the Agents SDK as a way to build agents that can inspect files, run commands, edit code, and work on longer tasks inside controlled environments.
Context
The update brings together several parts of the same vision: stronger models, real-time voice, agents with tools, browser interaction, image generation, and infrastructure for controlled execution.
Rather than a single isolated launch, the package points to a broader shift in how AI enters daily work. Interaction is moving beyond chat and into voice, browser sessions, files, code, images, and long-running workflows.
Why it matters
- Real-time voice and live translation make AI more useful for support, training, meetings, and operational assistance.
- Codex in Chrome brings agents closer to the systems where work actually happens.
- The Agents SDK reinforces the idea of agents working in controlled environments with files, commands, and tools.
- For companies, this is strong demo material for the next generation of AI workflows.