BODY: OpenAI has pulled back the curtain on GPT-5.6, and this time the company is splitting its newest model into a trio of distinct personalities β each named after a celestial or earthly body, and each pitched at a different kind of user.
At the top sits Sol, the flagship. It's the heavyweight built for the most demanding reasoning, coding, and research workloads β the model you reach for when raw capability matters more than the bill. Beneath it, Terra plays the role of the dependable middle child: a balanced option that trades a sliver of peak performance for broader everyday usability. And rounding out the lineup is Luna, engineered for speed and low cost, aimed squarely at high-volume, latency-sensitive deployments where every millisecond and every cent counts.
The rollout itself carries an unusual wrinkle. Rather than launching straight to the public, GPT-5.6 is debuting as a limited preview at the request of the US government β a detail that hints at the deepening entanglement between frontier AI labs and Washington. General availability is slated to follow once that preview phase concludes.
The three-tier structure mirrors a broader industry shift away from one-size-fits-all models and toward menus of options, letting developers dial in the precise trade-off between intelligence, latency, and price for each task.
The insider take
From Tokyo, the government-first preview is the line that jumps out. Japanese enterprises and public agencies have grown increasingly cautious about where their AI compute and data flows, and a model that ships to US officials before reaching the open market reinforces a quiet anxiety here: that access to the very best tools may increasingly hinge on geopolitics, not just budget. Expect Japanese firms to watch the Terra and Luna tiers closely β the pragmatic, cost-aware options tend to win in a market that prizes reliability and lean operations over bragging rights.
Originally reported by γ―γ¦γͺγγγ―γγΌγ― (Japanese).