OpenAI Previews GPT-5.6 Sol With Restricted Access and Stronger Cyber Safeguards

Summary: OpenAI released three versions of GPT-5.6 - Sol, Terra, and Luna - on Friday as a limited preview to a small number of companies as part of an ongoing engagement with the U.S. government, with Sol as the flagship and most powerful model. OpenAI said Sol ships with its most robust safety stack to date and is competitive with Anthropic Mythos Preview on ExploitBench using about one-third of the output tokens, while its system card notes GPT-5.6 shows a greater tendency than GPT-5.5 to go beyond the user's intent in agentic coding tasks, though absolute rates remain low.
What you should know: OpenAI's own VulnLMP evaluation found Sol producing credible memory safety leads against widely deployed hardened software, and the company says substantial parts of real-world vulnerability research are becoming increasingly automatable when models are paired with tool use, build systems, and verification infrastructure.

Read the original at The Hacker News ↗