On October 7, OpenAI did something it has rarely done in the past: it made its latest flagship model directly available to free users.
As GPT-6 rolled out to everyone, ChatGPT entered a new model cycle in earnest. Paying users get GPT-6 Sol, while free users receive GPT-6 Luna. At the same time, ChatGPT added “Intelligent UI” capabilities: AI no longer merely outputs text, but can generate charts, interactive components, calculators, forms, and other content based on the task.
Just a few hours before GPT-6 was released, Anthropic got ahead of the launch by releasing Claude Haiku 5.5, also a small model, at one-tenth the price of its predecessor.
However, some developers also reported that GPT-6 Sol’s overall performance was not even as good as GPT-5.6 Sol’s. In the overall intelligence rankings by Artificial Intelligence, GPT-6 Sol and GPT-5.6 Sol received similar scores, while GPT-6 Luna ranked behind DeepSeek V4.1flash and Claude Haiku 5.5.
Meanwhile, Google’s Gemini 4 Argon, Mistral’s trillion-parameter model Mistral Large 4, and Beam from Reflection AI, which is backed by Nvidia, all made their debuts in rapid succession during the same week.
Industry observers compared the intensity of this past week’s AI releases to “New Year’s in America.”
Behind this highly publicized competition, new questions are also emerging. If everyone can use the most powerful models, where exactly do AI companies’ moats lie? The answer may not be the models themselves, but rather access, costs, and ecosystems.