OpenAI has finally launched an agent service that goes beyond simple conversation to perform actual collaboration: ChatGPT Work. You can now freely switch between the existing Chat mode, Work mode, and Codex mode within the desktop app. It runs on the newly released GPT-5.6 model family, and the speed at which it plans and handles complex tasks is noticeably different! 😄 It's also become much more accessible, as even free users can immediately try out the balanced Terra-based Work mode.

The most exciting parts of this update are the execution settings, Max and Ultra modes, as well as the Sites feature. When Max is enabled, the model engages in deeper, more extended reasoning on a single thread, while Ultra mode lets the main agent control up to four sub-agents in parallel to resolve complex tasks. Furthermore, the Sites feature—which creates a web app and deploys it to a Cloudflare-based subdomain with just a single prompt—has entered public beta. Simply calling @Sites within Work mode instantly generates your own personalized dashboard or work tracker page.

We’ve moved past the chatbot stage of simply asking and answering questions; it feels like having a real work partner that plans steps, analyzes files, performs real-time web actions, and delivers finished results. Since you can start using it just by updating the desktop app—no complex API integration or setup required—I highly recommend handing off your repetitive, tedious tasks like document summaries or data organization to Work mode starting today!