OpenAI released GPT-5.6 to the public on July 9, making its most capable language model broadly available. The Trump administration had requested government review before release, delaying the rollout by weeks.

The model comes in three variants: Luna for speed, Terra for balance, and Sol for maximum capability. Sol sets what OpenAI calls a new standard for both intelligence and efficiency in coding, knowledge work, and science.

What Changed from GPT-5

GPT-5.6 Sol outperforms previous frontier models while using fewer tokens and at lower estimated cost. OpenAI’s numbers show improvements in competitive benchmarks for reasoning, coding, and real-world problem-solving.

The three-tier approach lets different users pick different tradeoffs. Luna prioritizes speed for interactive tasks. Terra aims for everyday work. Sol takes the time to think harder.

Government Review Left Its Mark

The Trump administration made OpenAI submit the model for review 30 days before release. The Department of Commerce’s Center for AI Standards and Innovation tested it for cybersecurity risks. After additional testing, permission came through for public release.

This sets a precedent. Frontier AI models now face government scrutiny before becoming public. It slowed OpenAI’s rollout but didn’t block it—an important distinction.

Where This Lands

GPT-5.6 enters a market where Claude and Gemini already compete heavily. OpenAI still leads in adoption, but the gap has closed. Each release from each company now gets examined not just for capability but for cost per token and real-world performance.

The frontier is moving fast, and the market judges speed as much as ability.

FYI (keeping you in the loop)

Why does government review matter for AI models?

Large language models can be used to generate harmful content, bypass cybersecurity, or create misinformation at scale. Government review aims to catch safety risks before deployment. The tradeoff is slower product cycles and potential regulatory capture.