OpenAI has launched ChatGPT Work, a new AI agent that can complete multi-step tasks across connected workplace applications. The company also rolled out its latest frontier model, GPT-5.6, which powers the new offering.
ChatGPT Work is available on web and mobile for Pro, Enterprise, and Edu users, with Plus and Business subscribers set to receive access over the next few days.
ChatGPT Work can access connected applications such as Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, and CRMs to gather information, create documents, spreadsheets, presentations, and web apps, and carry out projects that can run for hours with limited user intervention.
The company said GPT-5.6 powers ChatGPT Work, improving its ability to reason through complex tasks, follow templates, and reference files while producing finished work products.
“ChatGPT Work is an agent in ChatGPT that helps you take on more ambitious tasks,” OpenAI said in its blog post. “It can gather information across your apps and workflows to create finished materials like sheets, slides, docs, and web apps, and stay with complex projects for hours by breaking them into smaller steps and completing them independently.”
The launch builds on Codex, OpenAI’s coding agent, which the company said is now used by more than 5 million people every week. While Codex initially targeted software developers, OpenAI said more than 1 million users now rely on it for work outside software development.
The company also announced that Scheduled Tasks can now automate recurring workflows. Users can ask ChatGPT to review Slack updates, monitor websites, summarise changes, update presentations from email feedback, or refresh meeting agendas on a recurring basis. Users can track progress, modify tasks, and approve important actions when required.
On desktop, ChatGPT Work introduces a built-in browser that allows the AI agent to access websites, online tools, and files from Google Workspace and Microsoft 365 within the application. OpenAI also expanded Computer Use capabilities, allowing ChatGPT to interact with local applications, browsers, and files to complete tasks in the background.
OpenAI said the standalone Codex app will merge into the new ChatGPT desktop app. The combined application adds capabilities including inline code editing, pull request reviews, support for multiple repositories, and faster computer use powered by GPT-5.6.
The existing ChatGPT desktop application will be renamed ChatGPT Classic.
The updated app is available globally for Windows and Mac users, while Chat, Work and Codex features will be available across all plans, including the Free tier.
The company is also launching Sites in public beta, allowing users to build and share interactive websites and web applications directly from ChatGPT. The feature can be used to create dashboards, project trackers, prototypes, reports, and internal portals that update as underlying information changes.
To support enterprise adoption, ChatGPT Work includes security and governance controls inherited from ChatGPT Enterprise. Administrators can manage plugin access, connected tools, browser permissions, and workspace policies. OpenAI also introduced an Auto-review system that checks important actions involving connected tools and APIs before execution.
According to OpenAI, nearly all internal teams, including finance and sales, already use ChatGPT Work and Codex.
GPT 5.6 is Here
OpenAI has finally launched GPT-5.6, after delaying its rollout at the US government’s request to assess security concerns. It has introduced a new family of AI models comprising Sol, Terra, and Luna, as the company seeks to improve coding, scientific reasoning, cybersecurity and enterprise AI workloads while offering different performance and cost trade-offs.
OpenAI has rolled out GPT-5.6 across ChatGPT, Codex and the OpenAI API, with availability beginning today.
GPT-5.6 Sol is OpenAI’s flagship model and the company’s most capable model to date. It can do complex tasks requiring advanced reasoning, coding, biology research, and cybersecurity. Terra is positioned as a lower-cost alternative for production workloads, while Luna is the fastest and most cost-efficient model in the family.
Together, the three models are intended to serve a wider range of developers and enterprises depending on performance and pricing requirements.
GPT-5.6 Sol is priced at $5 per million input tokens and $30 per million output tokens. Terra costs $2.50 per million input tokens and $15 per million output tokens, while Luna, the company’s lowest-cost model, is priced at $1 per million input tokens and $6 per million output tokens.
According to OpenAI, GPT-5.6 Sol sets a new state of the art on Terminal-Bench 2.1, a benchmark that measures command-line workflows involving planning, iteration, and tool coordination.
On Agents’ Last Exam, GPT-5.6 Sol model scored 53.6, outperforming Anthropic’s Claude Fable 5 (adaptive) by 13.1 points. At the medium reasoning setting, Sol exceeded Claude Fable 5 by 11.4 points while costing roughly one-quarter as much. OpenAI added that GPT-5.6 Terra and Luna also outperformed Claude Fable 5 on the benchmark at around one-sixteenth of the estimated cost.
In cybersecurity, GPT-5.6 Sol achieved a score of 73.5% on ExploitBench, a benchmark that measures a model’s ability to progress from identifying vulnerable code to arbitrary code execution, compared with 47.9% for GPT-5.5 at a comparable output token budget.
The company also reported gains on coding agent tasks. GPT-5.6 Sol, with its maximum reasoning mode, outperformed Claude Fable 5 while using 54% fewer output tokens.
OpenAI said the advantage extends across the model family, with GPT-5.6 Terra narrowly surpassing Claude Fable 5 and GPT-5.6 Luna outperforming Claude Opus 4.8 at their highest-scoring configurations. According to the company, both Terra and Luna achieved these results at roughly one-fifth of the estimated API cost and in about one-third of the time.
ALSO READ: Alteryx Inspire 2026: Three Questions Every Data Leader Should Take to Orlando
