On August 6, 2026, OpenAI rolled out an update for GPT-5.6 Sol in ChatGPT. The model provides more concise responses, avoids unnecessary formatting, and, according to OpenAI, produces about 68 percent fewer factual errors than the previous GPT-5.5 Instant. Another new feature is a slider that lets you set how long ChatGPT should take to think for each response. For free users, GPT-5.6 Luna is now the default model, and they also get unlimited text chats.
Key Takeaways
- GPT-5.6 Sol provides more direct responses and, according to OpenAI, has 68 percent fewer factual errors than GPT-5.5 Instant.
- A new slider replaces the distinction between Instant and Thinking modes.
- Free and Go users get GPT-5.6 Luna as the default, along with unlimited text chats.
- File uploads, image analysis, and other tools remain limited in the free plan.
What GPT-5.6 Sol Does Better Now
Sol is the most powerful model in the GPT-5.6 series and is reserved for Plus and Pro subscriptions. OpenAI has primarily tweaked two aspects: factual accuracy and response style. Instead of long lists with subheadings, Sol is now designed to get straight to the point and omit details that nobody asked for. Anyone who’s ever asked a simple question and received a three-page guide in response knows exactly what this is aimed at.
The second change is the aforementioned slider. Previously, you had to decide in advance whether ChatGPT would respond immediately or enter “Thinking Mode.” That distinction is now gone. Instead, you can continuously adjust the level of depth, depending on whether you’re looking up a phone number or trying to clarify a tax question. According to The Decoder, there are five levels available for this.
In ChatGPT Work and Codex, Sol remains unchanged for now. The update therefore only affects the classic ChatGPT interface.
GPT-5.6 Luna Becomes the Default for Free Users
There are significantly more changes to the free plan. Free and Go users will receive GPT-5.6 Luna as the default model later this week, replacing GPT-5.5 Instant. Luna is the smallest and fastest model in the series. OpenAI had recently slashed API prices for this model, as we described in our article on the new GPT-5.6 pricing. It is precisely this cost structure that makes the next step possible in the first place.
Starting next week, OpenAI will completely lift the message limits for text chats. Free users will then be able to type as much as they want, limited only by safeguards against abuse. There will also be a “Think” button for questions where Luna needs more time to think.
What won’t be unlimited: file uploads, image generation, and the other tools will retain their current limits. The free flat rate applies exclusively to text-only conversations.
Fewer factual errors, but the numbers come from OpenAI
OpenAI tested both models with prompts from the fields of finance, medicine, and law—areas where a fabricated answer could cause real harm. Here’s an overview of the results:
| Model | Available for | Factual errors compared to GPT-5.5 Instant |
|---|---|---|
| GPT-5.6 Sol | Plus, Pro | about 68 percent fewer |
| GPT-5.6 Luna | Free, Go | about 62 percent less |
These figures are based on internal measurements and have not yet been independently verified. This doesn’t eliminate hallucinations—it just makes them less frequent. For medical or legal questions, a second source is still required.
The catch: The Think button doesn’t make Luna smarter
The Think button sounds like an upgrade, but it isn’t one in the true sense. It lets Luna compute longer, but it doesn’t switch to a more powerful model. Free users are therefore permanently stuck with the smallest member of the family, and smaller models are simply more prone to errors when handling complex tasks.
Previously, free users could at least occasionally access more powerful models. OpenAI is now closing that door and replacing it with unlimited usage. Whether this is a good deal depends on what you use ChatGPT for: For everyday questions, draft texts, and quick research, Luna is perfectly adequate. For anything that requires real thinking, there’s no way around a subscription.
OpenAI has also tightened its youth protection measures. Romantic role-playing games are blocked for users under 18, and the filters for sensitive topics like eating disorders are now stricter.
Rollout: What’s Coming and When
| Date | New Feature | For whom |
|---|---|---|
| Available now | Improved GPT-5.6 Sol and Reasoning controls | Plus, Pro |
| This week | GPT-5.6 Luna as the default model | Free, Go |
| Next Week | Unlimited text chats and Think button | Free, Go |
The changes apply across all platforms—on the web as well as in the apps for iOS, iPadOS, and macOS.
Conclusion
The update makes it pretty clear where OpenAI is headed: quantity for everyone, quality for a price. Unlimited text chats are a strong selling point against competitors who continue to cap their free offerings or, like Google with Gemini Spark, exclude entire regions. At the same time, the line between free and paid is becoming sharper than before.
For most casual users, this is a win. However, anyone using ChatGPT for work where mistakes can be costly shouldn’t confuse the 68 percent accuracy rate with error-free performance. And anyone publishing AI-generated content has been required to comply with the AI labeling mandate since August 2, regardless of which model is used.
