OpenAI began a limited preview of the GPT-5.6 series, including Sol, its flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable option. Terra provides performance comparable to GPT-5.5 at half the cost, while Luna offers strong capabilities at the lowest price. GPT-5.6 Sol launches with the company’s most robust safety features, including enhanced protections for high-risk activities and repeated misuse. These measures were developed through extensive testing and hardening against real-world attacks. Source: openai
GPT-5.6 Sol is the company’s most capable model to date, with improved agentic capabilities in coding, biology, and cybersecurity. It sets a new state of the art on Terminal-Bench 2.1, which evaluates command-line workflows requiring planning, iteration, and tool coordination. In biology workflows, it achieves stronger results than GPT-5.5 while using fewer tokens. For cybersecurity, it shifts the performance-efficiency frontier for long-horizon tasks, including vulnerability research and exploitation. On ExploitBench², GPT-5.6 Sol is competitive with Mythos Preview using only about one-third of the output tokens. Source: openai
The preview includes a layered safeguard stack designed to address potential misuse, with configurations tailored to each model’s capabilities. These safeguards include model-level refusal of prohibited cyber assistance, real-time misuse classifiers, account-level review, and differentiated access. OpenAI emphasized that while the model can identify bugs and exploitation primitives, it does not autonomously produce a functional full-chain exploit under tested conditions. The company is pairing increased capabilities with stronger safeguards and a phased release to ensure broader availability while managing risks. Source: openai
