On August 6, OpenAI announced an update that reworks everyday conversations in ChatGPT. For Plus and Pro users, GPT-5.6 Sol has been retuned for factual accuracy and tighter answers, and a new slider lets you choose how much thought goes into each response. Free users get GPT-5.6 Luna as their default model, along with unlimited text chats[1].
Fewer Factual Slips, Shorter Answers
The model updated here is the GPT-5.6 Sol that powers the Chat experience in ChatGPT. OpenAI says it has been tuned to make fewer mistakes when answers hinge on details such as dates, numbers, sources, rules, or assumptions, and to make better use of the sources it finds when assembling a response[1].
There are numbers behind the claim. In an internal evaluation using financial, medical, and legal prompts that demand factual detail, responses containing at least one factual error were about 62 percent less common with GPT-5.6 Luna and about 68 percent less common with GPT-5.6 Sol than with GPT-5.5 Instant[1].
The style of the output changes too. Answers are more focused, the level of detail adapts to the question, and unnecessary formatting is dropped. When simply agreeing would not be useful, the model may offer a gentle correction instead[1]. A quick question gets a direct answer with the context you need, while heavier work such as multi-step planning, research, or writing gets a fuller response that keeps the main recommendation visible.
A Slider for How Much the Model Thinks
Plus and Pro users get a slider that sets how much thought goes into an answer. It works in ChatGPT on web, mobile, and desktop, so you can keep everyday questions light and turn the dial up for planning, research, writing, coding, or decisions that need more consideration[1].
The less visible change matters just as much. For Plus and Pro, a single model now powers both Instant responses and deeper reasoning. Switching modes used to change the tone and writing style, which felt like handing the conversation to a different model. The point of the merge is to make higher effort feel like the same assistant taking more time[1].
Note that this version of GPT-5.6 Sol is optimized for everyday chats and is available only in the Chat experience. The GPT-5.6 Sol that powers ChatGPT Work and Codex is not part of this release[1].
Free Users Move to GPT-5.6 Luna with Unlimited Text
GPT-5.6 Luna becomes the default model for Free and Go users this week. Starting the following week, the cap on text chats is removed and a new Think button lets the model spend longer on harder questions[1].
Unlimited applies to text exchanges only. File uploads, images, and other tools keep their limits, and the unlimited tier is subject to abuse guardrails[1]. Still, the biggest friction in free usage, hitting a cap and waiting hours for it to reset, disappears for text conversations. Given that OpenAI says 1 billion people turn to ChatGPT every week, the number of people affected is substantial[1].
New Safeguards for Users Under 18
The release also covers safety training and evaluation. For users believed to be under 18, the model has been trained to avoid romantic roleplay, age-restricted challenges, and presenting itself as a substitute for real-world relationships[1].
OpenAI has also applied age-appropriate boundaries around sexual content, eating disorders and body-image risks, age-restricted goods, dangerous activities, and graphic violence, and the model encourages teens who may need support to connect with people they trust. This training is reinforced with system-level protections, and new evaluations have been added to measure how the models behave for users under 18[1]. Details are published in the system card.
Read Another Way, Free Users Are Pinned to the Weaker Model
The announcement has not been received uniformly. The outlet the-decoder framed the change as OpenAI restricting free users to its weakest model[2], since what reaches the free tier is GPT-5.6 Luna while the improved GPT-5.6 Sol stays behind the paywall.
You can read it as a trade: unlimited volume in exchange for a firm ceiling on which model free users can touch. The Think button extends reasoning time, but Luna is still the model doing the work. Free usage clearly got roomier, yet the line between free and paid has shifted from quota to the model generation itself.
Summary
On August 6, OpenAI updated the GPT-5.6 Sol that powers the Chat experience in ChatGPT, improving factual accuracy and answer concision. Plus and Pro users can use the thinking slider and the merged Instant-Thinking behavior starting the same day. Free and Go users receive GPT-5.6 Luna as the default model this week, with unlimited text chats and the Think button arriving the following week. In internal evaluations, responses containing factual errors were about 62 percent less common with Luna and about 68 percent less common with Sol compared with GPT-5.5 Instant. Removing the free tier cap is welcome, but the improved Sol remains a paid feature, and the free-versus-paid divide is moving toward the model generation itself.
Source: https://openai.com/index/improving-gpt-5-6-sol-in-chatgpt
