#OpenAI发布GPT-5.6 When the “Strongest Model” Meets the “Narrowest Gate”: GPT-5.6’s Limited Release

On June 27, Beijing time, OpenAI officially released the GPT-5.6 series of models. Unlike in the past, this time there were no lights or cheers—what followed instead was an unprecedented threshold. At the request of the U.S. government, the models are only available for preview to a small number of “trusted partners.”
GPT-5.6 is a model family with three tiers: the flagship Sol (sun), which set a new record with a 91.9% score in the Terminal-Bench 2.1 programming test; the balanced Terra (earth), whose performance benchmarks against GPT-5.5 while cutting costs by 2x; and the lightweight Luna (moon), focused on extreme cost-effectiveness. Sol introduces a “maximum reasoning intensity” mode and an “Ultra” mode that calls sub-agents, significantly improving efficiency on long-chain tasks such as network security and biology.
On pricing, Sol costs just half of competing product Claude Fable 5. However, the biggest signal of this release is this: the rules of the AI race have changed. When model capabilities reach national security boundaries, prior regulatory access becomes the new normal. GPT-5.6’s “narrow-gate” style release marks that frontier AI is evolving from public infrastructure into a strategic resource tightly controlled by geopolitics.