OpenAI Debuts GPT-5.6 Sol Debuts as Most Advanced AI Model Yet

OpenAI has announced the limited preview of its new GPT-5.6 series, introducing three models designed to serve different user needs.

The models include Sol, its flagship model; Terra, a balanced model for everyday tasks, and Luna, a fast, cost-effective model.

According to the company, Terra delivers performance comparable to GPT-5.5 at half the cost, while Luna offers strong capabilities at its lowest price point to date.

The company disclosed that GPT-5.6 Sol launches with its most advanced safety framework yet, featuring enhanced protections against higher-risk activities, sensitive cybersecurity-related requests, and repeated misuse.

OpenAI noted that it spent several weeks identifying vulnerabilities, conducting extensive stress tests, and strengthening the model’s defenses against real-world attacks before its release.

The AI company reiterated its commitment to broad access and revealed plans to make GPT-5.6 Sol, Terra, and Luna generally available in the coming weeks. However, the company explained that the rollout will begin with a limited preview involving a small group of trusted partners, following discussions with the U.S. government.

According to OpenAI, the participating organizations were disclosed to government officials as part of the preview process.

During the limited release period, OpenAI said it will continue testing the models and working closely with its partners to prepare for broader availability.

While acknowledging the government’s request for an initial restricted rollout, the company emphasized that it does not believe such an access process should become the long-term standard, arguing that it delays access to advanced AI tools for developers, enterprises, cybersecurity professionals, and users worldwide.

It stated that the temporary preview is intended to support the development of a cybersecurity Executive Order framework and establish a repeatable process for future AI model releases.

On the capabilities front, OpenAI described GPT-5.6 Sol as its most powerful model to date. Early evaluation results indicate improved agentic performance across coding, biology, and cybersecurity tasks, with additional safety and preparedness assessments documented in its system card.

The company said a broader set of benchmark results will be published when the models become generally available. It also introduced a new “max reasoning effort” mode, allowing GPT-5.6 Sol additional time to perform deeper reasoning on complex tasks.

In addition, OpenAI unveiled an “ultra mode” that extends beyond the capabilities of a single AI agent by deploying multiple subagents to collaborate on sophisticated workflows and accelerate problem-solving.

For software development, OpenAI reported that GPT-5.6 Sol has achieved a new state-of-the-art score on Terminal-Bench 2.1, a benchmark that evaluates command-line workflows requiring planning, iterative execution, and coordinated tool use, highlighting the model’s advancements in complex coding environments.

“GPT 5.6 Sol is our most capable model yet for cybersecurity. It shifts the performance-efficiency frontier for long-horizon security tasks including vulnerability research and exploitation”, the company wrote.

GPT 5.6 is trained to refuse prohibited cyber assistance, including when users attempt to disguise their intent or jailbreak the model. These model-level safeguards establish the first boundary around what the model should and should not help with.

The limited preview of the GPT-5.6 series signals OpenAI’s continued push toward building more capable, reliable, and commercially accessible AI systems.

As the company gathers feedback from trusted partners and completes additional safety evaluations, the broader rollout of Sol, Terra, and Luna is expected to provide developers, enterprises, and consumers with a wider range of AI options tailored to different performance and cost requirements.