OpenAI
GPT-5.6 Sol: half the errors
🧒 Discovery
ChatGPT's brain just got an upgrade: it makes far fewer mistakes than before.
📰 Essential
OpenAI rolls out GPT-5.6 Sol, cutting factual errors by 68% and improving reasoning for paid subscribers.
🔬 Expert
The 68% figure is OpenAI's own measurement against GPT-5.5, its previous model, on internal test sets: no independent public scores accompany the announcement, which calls for caution. The rollout follows the now familiar two-tier pattern, Sol for paying subscribers and an unlimited Luna variant for free users, and targets professional work first, where factual mistakes are costly.