OpenAI has launched ChatGPT Work, a new AI agent designed to help professionals complete complex workplace tasks without needing to know how to code.
Unveiled on Thursday, July 9, the product combines ChatGPT with Codex, OpenAI’s AI coding tool. Users can ask it to create documents, presentations, spreadsheets and websites, while the system handles much of the technical work behind them.
ChatGPT Work is powered by GPT-5.6, OpenAI’s latest family of AI models, which also launched on Thursday. The release gives OpenAI a direct competitor to Anthropic’s Claude Cowork as both companies fight for a larger share of the enterprise AI market.
What Is ChatGPT Work?
ChatGPT Work is an AI agent built for longer and more complicated professional projects.
Rather than asking users to manage each individual step, it can plan and complete multi-stage tasks based on the outcome they provide. This could include gathering information, working with files and connected applications, then turning that material into a finished presentation or report.
OpenAI said GPT-5.6 allows the product to reason through these tasks while following templates and using reference files supplied by the user. The goal is to give people who aren’t programmers access to the same execution capabilities that made AI coding tools increasingly useful to developers.
“You can apply the model’s ability to code to solve problems across every industry,” Ty Geri, product manager for ChatGPT Work, told Reuters.
OpenAI has also introduced a new ChatGPT desktop application that brings Chat, Work and Codex together. A hosted websites feature will allow users to build and share websites from inside ChatGPT Work rather than moving the project to a separate development platform.
The service began rolling out across web and mobile on Thursday for ChatGPT Pro, Enterprise and Edu users. OpenAI plans to expand access to Plus and Business subscribers over the following few days.
GPT-5.6 Arrives in Three Sizes
ChatGPT Work launches alongside three versions of GPT-5.6.
Sol is OpenAI’s flagship model, while Terra is positioned as a lower-cost option. Luna is the smallest, fastest and least expensive model in the family. This structure gives businesses a way to choose between maximum performance and lower operating costs, depending on the task they need the model to complete.
OpenAI said the new model family performs strongly across professional analysis, web browsing, tool use and computer-based tasks. The company is also placing particular emphasis on efficiency, arguing that the models can complete demanding work while using fewer resources than some competing systems.
Geri described GPT-5.6 as competitive with models that cost far more, while offering twice the speed at a lower price.
Max Weinbach, an analyst at Creative Strategies, told Reuters that the smallest version could complete some tasks about as well as the largest model at one-fifth of the cost.
“This is the first time where I’ve seen the small models complete these kinds of tasks,” he said.
The public launch of GPT-5.6 followed a delay requested by the US government over concerns that advanced AI models could be misused for cyber attacks or other activities affecting national security. Access was initially limited to selected partners while officials assessed the risks associated with a wider release.
When AI Drives Cyber Strategy
Infosec leaders rethink defence as agentic AI, autonomy and quantum risk force a move from basic protection to resilient, preemptive security.
OpenAI Takes Aim at Claude Cowork
ChatGPT Work puts OpenAI in more direct competition with Anthropic’s Claude Cowork.
Anthropic introduced Cowork in January as a way to bring the execution capabilities of its Claude Code product to people without programming experience. The agent can plan and complete longer tasks, continue working when a user closes their laptop and produce finished documents, spreadsheets and presentations for review.
The similarities between the two products reflect a wider move away from AI chatbots that mainly answer questions. OpenAI and Anthropic are now building systems that can take instructions, use software and complete more of the work themselves.
OpenAI previously offered several agent-based products, including Operator and deep research. These were later brought together as ChatGPT Agent for individual users, while Workspace Agents targeted enterprise workflow automation.
ChatGPT Work now gives the company a broader workplace product that connects those capabilities with Codex and GPT-5.6.
The launch also arrives as OpenAI and Anthropic compete for enterprise customers, where longer contracts and company-wide deployments can generate more revenue than consumer subscriptions. Both companies have been expanding their workplace products as businesses look for AI systems capable of completing complex tasks with less step-by-step input.
With ChatGPT Work, OpenAI is betting that wider availability, lower costs and a familiar ChatGPT interface will help it compete. Anthropic got there first with Claude Cowork. OpenAI now has its answer.
Comments ( 0 )