METR Publishes Predeployment Evaluation of GPT-5.6 Sol — Independent Read on a Government-Gated Frontier Model

METR·medium signal

Independent evaluator METR released its predeployment assessment of GPT-5.6 Sol on June 26, timed to OpenAI's limited preview and the model's High-risk Cybersecurity and Biological/Chemical designations. The report matters because it offers outside scrutiny of autonomous and agentic capabilities for a model whose public release is being directly gated by the U.S. government. It pairs with OpenAI's own GPT-5.6 Preview System Card on the Deployment Safety Hub, giving practitioners both first-party and third-party safety reads on launch day.