OPENAI
Meet Sol, Terra and Luna
OpenAI has started a limited preview of GPT-5.6, its next model family, with three versions: Sol, the most powerful one; Terra, the more balanced everyday model; and Luna, the faster, cheaper option.
The company says GPT-5.6 Sol is its strongest model yet, with better performance across coding, biology and cybersecurity.
It also adds a new max reasoning effort for deeper thinking, plus an ultra mode that uses subagents to handle more complex tasks.
Basically, one AI brain wasn’t enough, so now it has backup dancers.
Cybersecurity is a big focus.
OpenAI says Sol is better at helping users find and fix vulnerabilities, but it did not pass its “Cyber Critical” threshold, meaning it did not fully carry out an end-to-end cyberattack in testing.
Because the models are stronger in risky areas, OpenAI is rolling them out with extra safeguards.
These include built-in refusals, real-time checks for cyber and biology misuse, account-level reviews, and different access levels depending on risk.
The aim is to support legitimate work like code review, patching and defensive testing, while making harmful use harder.
The key bits:
Three models: Sol for power, Terra for balance, Luna for speed and lower cost.
Stronger cyber skills: useful for defenders, but wrapped in heavier safeguards.
Limited rollout for now: broader access is expected in the coming weeks.
Three new AIs, one velvet rope
The preview is currently limited to trusted partners through the API and Codex, with wider access planned soon for ChatGPT, Codex and API users.
OpenAI also says it previewed the release with the U.S. government before launch, while it works on a longer-term cyber safety framework.
Pricing starts at $5 input / $30 output per 1M tokens for Sol, $2.50 / $15 for Terra, and $1 / $6 for Luna.
OpenAI also plans to launch GPT-5.6 Sol on Cerebras in July, offering speeds of up to 750 tokens per second for select customers.
Whoever came up with the name, hats off to my girl/guy! - MG