OpenAI releases powerful new GPT-5.6 model but your access depends on government

OpenAI has announced GPT-5.6, its latest family of frontier AI models, but most users will have to wait before they can try it. According to the company, the new models are being released through a limited preview available only to a small group of trusted partners, following a request from the U.S. government. OpenAI said the restriction is temporary and that broader availability is expected in the coming weeks.

The GPT-5.6 release consists of three models named Sol, Terra and Luna, replacing OpenAI's previous version numbering with names inspired by celestial bodies. Sol is the flagship model designed for the most demanding workloads, Terra targets a balance between capability and efficiency, while Luna is intended for lower-cost and faster deployments.

According to OpenAI, the new models improve performance across coding, scientific research, cybersecurity and long-running agentic tasks. The company also introduced higher capability operating modes for Sol, including Max for more intensive reasoning and Ultra for orchestrating complex multi-agent workflows.

The launch also marks a shift in OpenAI's branding. Instead of continuing with numerical updates such as GPT-5.5 and GPT-5.6, the company has adopted named model families beginning with Sol, Terra and Luna. OpenAI said the change reflects a broader product strategy as its models become increasingly specialised for different workloads rather than simple sequential upgrades.

OpenAI says GPT-5.6 delivers stronger performance than GPT-5.5 across several technical benchmarks, particularly in software engineering, scientific reasoning and cybersecurity tasks. The company also compared its flagship model with other frontier AI systems, reporting competitive or leading results in a number of internal evaluations. Those comparisons were published by OpenAI as part of its technical announcement and have not been independently verified.

Unlike previous GPT launches, GPT-5.6 is not immediately available to all ChatGPT users or API customers. OpenAI said it is complying with a U.S. government request to initially provide access only to a limited set of trusted partners while authorities establish a framework for evaluating advanced AI models with significant cybersecurity capabilities.

The company said this restricted rollout is not intended to become the norm and argued that developers, businesses and cyber defenders benefit from timely access to advanced AI systems. OpenAI added that it expects to expand availability once the preview period concludes.

The GPT-5.6 announcement comes only days after OpenAI revealed Jalapeño, its first custom AI chip developed in collaboration with Broadcom. The company said the chip is designed to power ChatGPT, Codex, its API services and future AI products, marking an expansion of OpenAI's strategy beyond software into AI infrastructure. Together, the chip launch and the debut of GPT-5.6 highlight OpenAI's continued investment in both the hardware and software needed to support increasingly capable AI systems.