BREAKING: OpenAI just dropped the limited preview of its new GPT 5.6 model suite: Sol, the flagship; Terra, a medium-ti…
Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-26
OpenAI releases a US government-gated limited preview of GPT-5.6, a three-model suite of flagship Sol, mid-tier Terra, and affordable Luna, with Sol advancing agentic capabilities and cybersecurity performance over GPT-5.5 at the same price point.
Appears in
Extraction
Topics: openaigpt-5.6model-releaseai-regulationagentic-ai
Claims
- OpenAI released three models: Sol (flagship), Terra (GPT-5.5-level performance at half the cost), and Luna (fast and affordable for high-volume workloads).
- The US government requested a limited trusted-partner preview before broader public access.
- GPT-5.6 Sol excels at agentic tasks involving planning, tool use, self-correction, and multi-step execution.
- OpenAI used over 700,000 A100-equivalent GPU hours for automated red-teaming against jailbreak attacks.
- Sol is OpenAI's strongest model for vulnerability research but did not autonomously produce a full-chain exploit in tested Chromium and Firefox environments.
Key quotes
The most revealing part is the release gate: OpenAI says the U.S. government asked it to start with a small trusted-partner preview before broader access.
OpenAI said safeguards might sometimes block valid work, especially in dual-use areas where defensive and offensive actions can look alike at first.
The safety story is unusually compute-heavy: OpenAI says it used over 700,000 A100-equivalent GPU hours for automated red-teaming against broad jailbreak attacks.