OpenAI has officially released its latest language model, GPT-5.6, along with a new feature called ChatGPT Work, which transforms the chatbot into a more capable AI agent. This advancement allows ChatGPT to autonomously handle complex, multi-step tasks using integrated applications, documents, and various data sources.
The ChatGPT Work mode effectively combines the functionality of the traditional ChatGPT with the capabilities of the AI agent Codex. While Codex primarily catered to developers, the new ChatGPT mode opens its technology to everyday users. It can now connect to platforms like Slack, Microsoft Teams, Google Drive, SharePoint, Gmail, and CRM systems, utilizing data from these applications to accomplish user-defined tasks.
OpenAI describes the new agent's ability to gather context from selected applications and workflows, enabling it to produce comprehensive materials such as documents, spreadsheets, presentations, and even web applications. ChatGPT Work can automatically select the most relevant information sources or query specific services based on user requests. Additionally, it includes a Scheduled Tasks feature, allowing actions to be triggered by specific timeframes or events.
In tandem with the new model, OpenAI has updated the ChatGPT desktop application for Windows and macOS, introducing a built-in browser for web service interactions and a Computer Use function that enables AI engagement with local applications and files. Another notable addition is the Sites feature, through which ChatGPT can create small websites and web applications based on user descriptions.
The foundation of these new capabilities lies in the GPT-5.6 model family, which consists of three versions: Sol, Terra, and Luna. Sol is designed for the most demanding tasks, Terra serves as a versatile option, and Luna focuses on speed and cost-effectiveness. According to OpenAI, GPT-5.6 offers significant improvements over its predecessor, GPT-5.5, particularly in programming, document analysis, and computer interface tasks, all while reducing computational costs.
OpenAI aims to establish the Sol model as the gold standard for intelligence and efficiency, especially in programming, cybersecurity, and scientific research, positioning it as a more cost-effective alternative to competitors' top models amid widespread financial constraints in the AI industry.
Furthermore, OpenAI has enhanced safety mechanisms for GPT-5.6, revamping its protection against potentially harmful requests. This includes real-time monitoring and additional checks for high-risk actions. The most sensitive features in cybersecurity will remain exclusive to participants in the Trusted Access program, who undergo additional vetting.
The rollout of GPT-5.6 and the ChatGPT Work mode has begun, with the new models available immediately to all users of the ChatGPT desktop app, including those with free accounts. Access to GPT-5.6 and the Work mode will first be granted to subscribers of Pro, Enterprise, and Edu plans in the web and mobile versions, with Plus and Business plan users receiving access in the coming days. The GPT-5.6 models will gradually be available through ChatGPT, Codex, and the OpenAI API, with usage costs calculated per million tokens based on the model version.
This launch signifies a substantial leap forward in AI capabilities, likely intensifying competition in the market as rivals strive to keep pace with OpenAI's advancements.
Informational material. 18+.
