Skip to main content
AI Socratic

@OpenAI announced the limited preview of its next-generation models: GPT-5.6 Sol (flagship), Terra (balanced & efficient), and Luna (fast & affordable).

They plan broad availability in the coming weeks, but for now it's restricted to a small group of trusted partners via Codex and the API — at the request of the U.S. government.

image.png

The lineup

  • Sol — Step-function improvement over GPT-5.5. New SOTA on Terminal-Bench 2.1 (complex CLI workflows) and particularly strong on cybersecurity tasks (vulnerability research & exploitation).
  • Terra — Competitive performance with GPT-5.5 at roughly 2x lower cost.
  • Luna — Strong capability at the lowest price point yet — optimized for high-volume work.

Together they give developers more granular choice across intelligence, speed, and cost.

A few thoughts

This is classic OpenAI product strategy: drop a capable frontier family with clear tiering while navigating government constraints. Sol looks like the direct competitor to Anthropic's Fable/Mythos-class models, while Terra and Luna put pressure on both closed and open-source offerings on price/performance.

The solar-system naming theme is memorable (and already triggering crypto PTSD with the Terra/Luna references). The real story is how quickly the ecosystem is fragmenting into capability tiers — and how government "trusted preview" gating is becoming the new normal for frontier releases.

Worth watching how quickly they expand access once the benchmark and safety discussions with the U.S. government conclude.

Sources: OpenAI announcement thread · Blog post

About the Authors

Federico Ulfo

Federico Ulfo

Founder, Engineer

New York City