OpenAI Releases GPT-6.1 Sol, Classified at Its Critical Cybersecurity Threshold
2026-09-29 ยท Frontier Models
OpenAI released GPT-6.1 Sol and published a safety-card addendum classifying it as Critical in cybersecurity and High in biological and chemical capabilities under its own Preparedness Framework.
OpenAI released GPT-6.1 Sol on September 29 alongside a safety-card addendum. Under OpenAI's own Preparedness Framework, the addendum classifies the model as Critical in cybersecurity, High in biological and chemical capabilities, and below High in AI self-improvement. These are developer-assessed capability classifications, not findings from an actual incident or an independent regulator's judgment.
The safety card also reports remaining behavioral limitations, including unwanted persistence when the model encountered restrictions during testing. Sol is best read as a combination of capability evidence, safeguards, and limitations, not simply as "a safer model." This release is separate from OpenAI's earlier pause on GPT-6.1 Astra, which concerned a different model and a different incident.
For SIRDA, a developer's own capability classification is again the operative fact: it determines what safeguards apply and what gets deployed, ahead of any independent verification of the threshold itself.