OpenAI just let GPT-5.6 out of the lab and into everyone's hands

GPT-5.6 is officially out of beta, and instead of one model, OpenAI has handed everyone three.

The rollout went fully global on 9 July after OpenAI had spent roughly two weeks restricting access to a small group of trusted partners at the request of the US government. Sol is the new flagship, Terra is the everyday workhorse, and Luna is the cheap, fast option for anyone who does not need the smartest model in the room for every single task. It is a genuine restructure, not just a version bump.

Here is what is actually different. Sol is setting new benchmarks across coding, knowledge work, cybersecurity and science, reportedly outperforming both previous OpenAI models and rival frontier models while using fewer tokens at a lower estimated cost. OpenAI has also introduced something called ultra, a new highest capability setting that coordinates multiple agents working in parallel to finish complex tasks faster than a single model working alone.

Access depends on your plan. Plus, Pro, Business and Enterprise users get Sol through medium and higher effort settings, while GPT-5.6 is available now across ChatGPT, Codex and the OpenAI API rather than being locked to one product.

The timing is not subtle. This drops the same fortnight Anthropic has been publicly backpedalling on Claude Fable 5 pricing, extending free access twice after backlash over pulling it from standard plans. When your closest rival is fumbling its own pricing rollout in real time, landing a cheaper, more capable model family the same week is about as loud a flex as this industry gets.

For anyone using ChatGPT for actual work rather than novelty, three tiers means fewer excuses to overpay for tasks that do not need premium brainpower. If your business has been paying flagship rates for things Luna or Terra could handle just as well, this is the week to check your setup.

The cynical read is that OpenAI wanted a headline before Anthropic's next move landed. The practical read is that frontier AI just got both more capable and noticeably cheaper to run properly in the same release, and that combination matters more to most people than another spec sheet ever will.