OpenAI Ships GPT-5.6 Series in Limited Preview
*OpenAI released three new models—Sol, Terra, and Luna—under a restricted rollout requested by the US government.*
The release
OpenAI placed its GPT-5.6 family into limited preview today. The lineup contains Sol, the top-tier model, Terra for routine work, and Luna, the low-cost option.
Terra matches GPT-5.5 performance at half the price. Luna delivers usable capability at the lowest rate OpenAI currently offers. Sol adds agent-style gains in coding, biology, and cybersecurity work.
New controls
The company introduced a “max” reasoning setting and an “ultra” mode that spawns sub-agents for harder tasks. It also claims the strongest safety measures yet, covering high-risk domains, sensitive cyber queries, and misuse attempts.
Testing included red-team exercises and hardening against observed attack patterns. The models remain behind preview access only.
Government role
OpenAI stated the limited preview follows a direct request from the US government. No further details on scope or duration were provided in the announcements.
Why it matters
Developers and teams that rely on OpenAI APIs now face another tiered pricing and capability ladder. The safety claims will require independent verification before production use in regulated fields. The government involvement signals that frontier-model releases are no longer purely commercial decisions.
---
Sources:
{
"excerpt": "OpenAI launched GPT-5.6 models Sol, Terra, and Luna in limited preview after a US government request.",
"suggestedSection": "ai",
"suggestedTags": ["openai", "gpt-5.6", "ai-models"],
"imagePrompt": "Abstract neural lattice suspended inside a reinforced vault, soft directional light catching metallic nodes against matte concrete walls. Muted color palette, cinematic lighting, 16:9.",
"imagePrompt": "Abstract neural lattice suspended inside a reinforced vault, soft directional light catching metallic nodes against matte concrete walls. Muted color palette, cinematic lighting, 16:9."
}
No comments yet