Hey everyone! OpenAI released GPT-6.1 Sol today, and Ultra AI in AtomAI is already running on it. If you use Ultra AI, the next hard question you ask is answered by GPT-6.1 Sol. There is nothing to switch on and nothing extra to pay.
Only Ultra AI has moved. Everyday questions on every plan still run on GPT-6 Luna, exactly as they did last week.
Close to Astra, at a fraction of the cost
GPT-6 Astra is OpenAI's largest model. OpenAI says GPT-6.1 Sol nearly matches it on agentic coding, computer use and professional work, at one fifth of Astra's price. That price is what lets us put it behind Ultra AI without changing your plan or your allowance.
On software engineering it edges past Astra, scoring 75.2% on DeepSWE against 68.8% for GPT-6 Sol. It also reads documents about as well as Astra does, which matters when you ask Ultra AI about a PDF or a long set of notes.
Fewer wrong answers
By OpenAI's measure, GPT-6 Sol got 11.4% of hard factual questions wrong at low reasoning effort. GPT-6.1 Sol gets 7.7% wrong. OpenAI also says it is more upfront about its limitations, so when it is not sure, it is more likely to tell you than to guess.
OpenAI reports gains in debugging code, understanding documents and working through multi-step tasks, which is the kind of question Ultra AI is for: a maths problem with several steps, a proof, a bug that will not go away.
One change to how thinking works
GPT-6.1 Sol always thinks before it answers, even briefly. So the Faster setting in Ultra AI now does a short burst of reasoning instead of none at all. It is still the quickest option. Balanced and Deep work as before, and Research mode runs on GPT-6.1 Sol too.
What you need to do
Nothing. Turn on Ultra AI and you are on GPT-6.1 Sol. Your chats, files and study sets are where you left them, and your plan and usage allowances have not changed. Free accounts can try it with their free Ultra AI answers.
Log in and give GPT-6.1 Sol your hardest question!
You can read more about GPT-6.1 Sol in OpenAI's announcement here.
