GPT-5.6 and ChatGPT Work Agent: Automation Goes End-to-End

From Chatbot to Work Engine: What GPT-5.6 and ChatGPT Work Really Change

GPT-5.6 and the ChatGPT Work agent together mark OpenAI’s shift from conversational AI toward end-to-end enterprise automation, combining three tuned GPT-5.6 models with a desktop and cloud agent that can run multi-step tasks across business tools for hours without constant human supervision. This is not a quiet iteration; it is an explicit attempt to turn ChatGPT into an AI workflow tool that replaces dozens of manual steps in knowledge work. OpenAI launched GPT-5.6 on Thursday as a three-model family—Sol, Terra, and Luna—while also unveiling ChatGPT Work, a workplace agent that runs multi-step tasks across Slack, Google Drive, and Microsoft 365 and merges the Codex app into a rebuilt desktop client. These moves land after a government-mandated national security testing delay that pushed the release from June to July, and they are clearly calibrated to counter Anthropic’s momentum on price, speed, and productivity.

Three GPT-5.6 Models: Sol, Terra, Luna and a Pricing War

OpenAI did not ship one GPT-5.6; it shipped a tiered family meant to cover flagship reasoning, mainstream use, and high-volume workloads at lower cost. Sol is the top model, roughly equivalent to GPT-5.5’s thinking mode and aimed squarely at Anthropic’s Claude Fable 5. It scores 59 on the Artificial Analysis Intelligence Index, within one point of Fable 5’s 59.9, while costing roughly one-third as much per task. Pricing runs USD 5 (approx. RM23) per million input tokens and USD 30 (approx. RM138) per million output for Sol, USD 2.50 (approx. RM11.50) and USD 15 (approx. RM69) for Terra, and USD 1 (approx. RM4.60) and USD 6 (approx. RM27.60) for Luna. Terra sits at GPT-5.5-class capability at about half the cost, while Luna plays the instant-response role with the lowest price and capability tier. OpenAI is openly using these GPT-5.6 models as price-performance weapons against Anthropic and other frontier labs, backed by benchmark wins and cheaper tokens.

ChatGPT Work Agent: Multi-Step Task Automation Across Enterprise Tools

If GPT-5.6 is the engine, ChatGPT Work is the transmission that connects it to real work. The ChatGPT Work agent is a workplace tool that runs multi-step tasks across Slack, Google Drive, and Microsoft 365 and allows the AI to access your desktop and browser to perform tasks for you. It can create spreadsheets, presentations, and full websites, then keep going for hours by breaking complex projects into smaller tasks. In the same way that Anthropic’s coworking agent connects to outside applications, ChatGPT Work plugs into Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, CRMs, and project trackers, among other tools. This is no longer about generating a single document; it is about enterprise automation where an AI workflow tool can take charge of files, coordinate across apps, and push a project from input to finished outputs with minimal manual intervention. The unsettling demo where an OpenAI staffer let Work reorganize all his Apple Notes shows how much autonomy the agent now has over user data and file structures.

Productivity, Positioning, and a Desktop Land Grab

OpenAI’s intent with GPT-5.6 and ChatGPT Work is obvious: beat Anthropic on price, speed, and productivity, and own the desktop where knowledge work happens. The company’s announcement compares Sol directly to Fable 5 on multiple benchmarks, claiming near-parity in broad intelligence at much lower cost. It leads the Artificial Analysis Coding Agent Index at 80 points and hits 91.9% on Terminal-Bench 2.1, and internal partners report lower latency and fewer tokens per code review when using GPT‑5.6 versus GPT‑5.5. At the same time, the rebuilt ChatGPT desktop app adds a built-in browser, computer control capabilities, multi-tab support, and enterprise authentication, while folding the Codex engineering environment into a unified interface. The company is retiring legacy architectures—GPT-5.2 and GPT-4.5 have been fully shut down, and the o3 model will be deprecated by late August 2026. This consolidation, led by its applications chief, is a strategic bet that one AI surface tied to strong models and a capable work agent will become the default AI workflow tool for office workers.

What Enterprise Teams Should Do Next

For enterprises, the release is less a shiny upgrade and more a prompt to rethink process design around autonomous agents. Access to GPT-5.6 opens today across ChatGPT, Codex, and the API, with Sol available to Plus, Pro, Business, and Enterprise users, while Free and Go users get Terra; the rollout may take up to 24 hours. ChatGPT Work is rolling out on web and mobile for Pro, Enterprise, and Edu plans, with Plus and Business following over the next few days. According to OpenAI, the best way to learn the agent is to give it tasks you already know well—analyze a month-end budget variance, turn source materials into a marketing campaign brief, or prepare for a sales meeting. Internal testing claims nearly 100% of OpenAI’s teams now use ChatGPT Work, with finance teams cutting month-end close from days to hours. That is the opportunity: research, app building, and workflow automation are now ready to be delegated to an AI that can run for hours across your tools. The risk is trust—deciding which files and processes you are willing to let an autonomous agent touch. Teams that start with constrained, well-understood workflows will be best positioned to capture the productivity gains while keeping a human hand on the steering wheel.