이미지: OpenAI
Summary
- OpenAI improved GPT-5.6 Sol for Plus and Pro users, boosting response conciseness and factual accuracy while adding a slider to adjust thinking depth.
- The default model for Free and Go users is being switched to GPT-5.6 Luna, with unlimited text conversations and a 'Think' button arriving starting next week.
- Usage limits still apply to tools such as file uploads and image generation, and free users will not have access to frontier reasoning models.
What's Changed in GPT-5.6 Sol
OpenAI has updated GPT-5.6 Sol for ChatGPT Plus and Pro users. The update centers on two key changes. First, the response structure has changed so that simple questions get concise, to-the-point answers, while complex tasks get fuller responses with key recommendations clearly surfaced. Second, factual accuracy has improved. In internal evaluations using finance, medical, and legal prompts, OpenAI said GPT-5.6 Sol showed roughly a 68% reduction in responses containing factual errors compared to GPT-5.5 Instant. This figure has not been independently verified.
Previously, Instant and Thinking modes in ChatGPT operated with different tones and styles, making them feel like two separate models. With this update, OpenAI aims to unify the two experiences into a single, consistent model. Rather than switching to a different model when a user raises the slider, the same model spends more time to produce a more thorough answer.
Plus and Pro Users: Adjustable Thinking-Depth Slider
The change Plus and Pro users will notice most directly is the thinking-depth slider. On web, mobile, and desktop ChatGPT, setting the slider low delivers quick answers for everyday questions, while setting it high yields deeper responses for complex tasks like planning, research, writing, coding, and decision-making. The slider was already available in ChatGPT Work, and is now being introduced to regular ChatGPT as well.
However, the improved GPT-5.6 Sol released this time is optimized only for the Chat experience. The GPT-5.6 Sol used in ChatGPT Work and Codex is not part of this update and remains unchanged. In other words, Work and Codex users will not immediately receive these Sol improvements.
Free Users: Luna as Default, Unlimited Text on the Way
The default model for Free and Go users will be switched to GPT-5.6 Luna this week. GPT-5.6 Luna is the smallest model in the GPT-5.6 family, but OpenAI said it is an advancement over GPT-5.5 Instant, the previous default model. Unlimited text conversations, along with a 'Think' button that lets Luna spend more time on complex questions, will roll out gradually starting next week. The Think button is subject to anti-abuse guidelines.
According to OpenAI, GPT-5.6 Luna also showed about a 62% reduction in responses containing factual errors compared to GPT-5.5 Instant. However, the Think button only allows Luna to reason for longer — it does not switch users to a higher-tier model. As a result, free users still won't have access to OpenAI's frontier reasoning models.
Terms of Use and Limitations
Usage limits still apply to tools such as file uploads, image generation, and voice features. OpenAI did not disclose specific limit figures in this announcement. It's also worth noting that the factual accuracy improvements cited for the updated GPT-5.6 Sol come from OpenAI's internal evaluations and have not been independently verified by outside parties.
On the safety front, OpenAI said it has trained the models to avoid romantic roleplay, adult content, dangerous activities, and graphic violence for users believed to be under 18, with related details disclosed in the system card. This announcement does not include any changes to pricing or subscription fees.
