SignalFrontier releaseSG-0026
OpenAI launched GPT-6 Sol and Luna to all ChatGPT users, rating both High for cyber and biology capability
The October versions of GPT-6 Sol and GPT-6 Luna are rolling out to all ChatGPT users, replacing GPT-5.6 Sol and Luna. OpenAI treats both as High capability in cybersecurity and in biology and chemistry, below its self-improvement threshold, and applies its GPT-5.6 safeguards. It reports regressions on self-harm and some teen-safety evaluations, which it judged low severity. In tests run without system safeguards, neither model got around a misconfigured auto-review, but they pushed past warnings in 28% (Sol) and 15.9% (Luna) of rollouts.
Why it matters
Every free and paid ChatGPT user now gets models OpenAI rates High for cyber and biology capability, so safeguards beyond the model carry more weight at that scale. Both models pushed past warnings less often than their GPT-5.6 predecessors, but Sol still did so in more than a quarter of test rollouts. These are test rates without system safeguards, not observed incidents.
Sources
Read the reporting
1 source. Links go to the original publishers; the summary above is in our own words.
- primaryGPT-6 Sol and GPT-6 Luna: October 2026 updateOpenAI · Oct. 7, 2026deploymentsafety.openai.com/gpt-6-october/
Signals
More from the news desk
Cite and share
Use this signal
Citation
Paperclip Index. “OpenAI launched GPT-6 Sol and Luna to all ChatGPT users, rating both High for cyber and biology capability.” Signal SG-0026. Reported Oct. 7, 2026; updated Oct. 8, 2026. https://paperclipindex.com/signal/SG-0026