SAN FRANCISCO, CALIFORNIA / RankWire.AI / – OpenAI has released the GPT-5.6 model family and introduced ChatGPT Work, an agent designed for longer and more complex assignments. The family includes GPT-5.6 Sol, Terra and Luna. Each model targets a different balance of capability, speed and cost. OpenAI said the models improve performance across coding, research, professional work, computer use, science and design. The company has also expanded GPT-5.6 access across ChatGPT, Codex and its application programming interface.
GPT-5.6 Sol serves as the flagship model for difficult reasoning and professional tasks. Terra offers strong performance at a lower cost, while Luna focuses on speed and high-volume workloads. OpenAI has also introduced a new maximum reasoning setting for Sol. An ultra mode can use multiple subagents to divide complex work into separate tasks. The company said GPT-5.6 improves token efficiency and supports more advanced software development, scientific analysis and cybersecurity work.
In ChatGPT, GPT-5.6 Sol now powers the Medium, High and Extra High reasoning options on eligible plans. GPT-5.6 Sol Pro handles the most demanding tasks and longer-running workflows. GPT-5.5 Instant remains the default model for faster daily requests. Terra and Luna are not selectable in standard ChatGPT conversations. However, eligible users can access all three models through ChatGPT Work, Codex and the OpenAI API, depending on their plan and product access.
ChatGPT Work handles longer assignments across files and apps
ChatGPT Work can research information, analyze files and use connected workplace applications. It can also create documents, spreadsheets, presentations, reports and websites. Users can follow its progress, answer questions and change its direction while it works. Important actions still require user approval. The agent can gather context from connected files and tools before organizing a project into steps. OpenAI said GPT-5.6 powers the service and supports assignments that may continue for extended periods.
The Work agent also supports Scheduled Tasks for recurring or event-based assignments. Users can set a task to run once, repeat on a schedule or monitor selected information for changes. The system can then prepare reviewable work products inside ChatGPT. OpenAI has made Work available on desktop and is expanding access across supported plans and devices. Availability varies by subscription, product and region, including separate rules for standard ChatGPT conversations and other OpenAI services.
Developers gain new reasoning and workflow controls
For developers, the GPT-5.6 alias routes requests to GPT-5.6 Sol. OpenAI recommends Terra for workloads that need capable performance at a lower price. Luna supports cost-sensitive applications that require faster processing at scale. The family also adds programmatic tool calling, explicit prompt caching controls and persisted reasoning. Developers can use standard or pro reasoning modes in the Responses API. The models also support large context windows and long outputs for detailed production tasks.
OpenAI published updated safety information alongside the release. The company said it tested GPT-5.6 across cybersecurity, biological research and other dual-use areas. Its safeguards include model-level protections and additional controls for sensitive requests. OpenAI said GPT-5.6 Sol did not cross the critical cybersecurity threshold under its Preparedness Framework. The release gives businesses, developers and ChatGPT users a new set of models for complex work, while ChatGPT Work brings those capabilities into a managed agent workflow.