OpenAI launches ChatGPT Work and GPT-5.6 model

OpenAI has just announced two new product lines, ChatGPT Work and the GPT-5.6 model, marking the next step in its strategy of bringing AI into professional tasks and business processes. ChatGPT Work is a type of "agent task" within ChatGPT, designed to handle complex and lengthy tasks on behalf of users. According to OpenAI's description, ChatGPT Work can access and synthesize information from many applications and workflows, then create complete products such as spreadsheets, presentation slides, text documents or web apps, and at the same time divide large projects into many steps and complete them one after another in many hours. The platform is built on Codex, a code-based agent that OpenAI recently expanded to support productivity tasks, not just limited to software development. ChatGPT Work is operated on the GPT-5.6 Sol model, OpenAI's new frontier model, focusing on multi-step reasoning and content creation based on templates and reference files. The service begins rolling out today to users on ChatGPT Pro, Enterprise and Edu plans, and will soon be available on Plus and Business plans in the coming days. Notably, if using the ChatGPT desktop application, free plan users can also access ChatGPT Work. At the same time, OpenAI introduced the GPT-5.6 model set including Sol as the new flagship model, Terra for daily work and Luna as the most cost-effective model. As explained by OpenAI, GPT‑5.6 Sol sets a new benchmark in both intelligence and efficiency, achieving top results in programming, knowledge work, cybersecurity and scientific tasks, while surpassing previous frontier models and competitors with fewer tokens and lower costs. OpenAI also added ultra mode as the highest configuration, allowing the coordination of multiple agents in parallel to handle heavy tasks faster, emphasizing the ability to use computers and design evaluation capabilities to become a "smoother" collaborator, helping to test, refine and return ready-to-use results. In terms of safety, GPT‑5.6 Sol is equipped with OpenAI's strongest protection system to date and is described as the company's best programming model, surpassing Anthropic Fable 5 in coding tests. Terra is also said to surpass Fable 5, while Luna surpasses Opus 4.8, all three achieving higher performance at about two-thirds the cost. The GPT-5.6 generation also has improved interface design assessment, helping AI create “aesthetic, usable and practical” interfaces, along with upgrades in computing power, knowledge work, cybersecurity and research. GPT‑5.6 is now available across the OpenAI ecosystem, including ChatGPT, Codex, and APIs, but the specific implementation depends on the service plan and model chosen by the user. Regarding pricing, OpenAI applies a rate per 1 million tokens with three configurations: Sol costs 5 USD for input and 30 USD for output, Terra is 2.5 USD for input and 15 USD for output, and Luna is 1 USD for input and 6 USD for output. With this new step, OpenAI continues to promote its AI product group aimed at businesses and professional users, focusing on performance per cost and applicability in real work processes.