본문으로 건너뛰기
All news

OpenAI Launches GPT-5.6 as Sol, Terra, and Luna — Restricted to US Trusted Partners

Summary: OpenAI launched GPT-5.6 in three variants — Sol, Terra, and Luna — but complied with a Trump administration request to restrict initial access to vetted US partners while the government conducts a cybersecurity review.

Key Points

  • Sol: OpenAI's strongest model to date, leading on cybersecurity, coding, and biology tasks
  • Terra: A balanced, general-purpose model optimized for everyday use
  • Luna: A faster, lower-cost variant built for speed-sensitive applications
  • The White House asked for a staggered rollout to assess risks of cyberattack facilitation and military misuse before public release
  • OpenAI pushed back: "We don't believe this kind of government access process should become the long-term default"
  • GPT-5.6 includes a redesigned reward audit pipeline that addresses the alignment failure documented in GPT-5.5, and extends context to 1.5 million tokens

Why It Matters

Following Anthropic's Fable 5 export controls, a second major US frontier model has now faced pre-launch government review — a pattern rapidly becoming de facto policy. US labs operating under security constraints face mounting competitive pressure from Chinese counterparts who ship freely. Whether this becomes formal regulation or remains ad hoc White House pressure will define how frontier AI reaches the market.

Further Reading

뉴스레터 구독

무료 뉴스레터

매주 핵심 AI 소식, 한 번에 받기

쏟아지는 AI·LLM 뉴스 중 꼭 알아야 할 것만 골라 메일로 보내드려요. 뉴스레터 발송이 시작되면 구독자분들께 가장 먼저 보내드립니다.