OpenAI improves GPT-5.6 Sol in ChatGPT and restricts free users to its weakest model
Quick Answer
OpenAI has enhanced GPT-5.6 Sol for Plus and Pro users, offering a response depth slider, while free users will be limited to the less capable GPT-5.6 Luna.
Quick Take
The new models show improved factual accuracy, with error rates dropping 62% for Luna and 68% for Sol compared to GPT-5.5 Instant.
Key Points
- GPT-5.6 Sol offers focused responses and a depth adjustment slider for paying users.
- Free users will default to GPT-5.6 Luna, which is less capable than Sol.
- Factual error rates dropped by 62% for Luna and 68% for Sol compared to GPT-5.5 Instant.
- The new 'Think' button for Luna allows longer reasoning but lacks advanced capabilities.
- The update is exclusive to ChatGPT; GPT-5.6 Sol remains unchanged in Work and Codex.
DeepSignal Analysis
What happened
OpenAI has introduced enhancements to its GPT-5.6 Sol model for Plus and Pro users, including a response depth slider. Free users will be limited to the less capable GPT-5.6 Luna, which is set to become the default model for them. The company claims improved factual accuracy, with error rates decreasing by 62% for Luna and 68% for Sol compared to the previous GPT-5.5 Instant model.
Key evidence
- OpenAI has enhanced GPT-5.6 Sol for Plus and Pro users, introducing a response depth slider for adjusting the detail of responses.
- Free users will be limited to the GPT-5.6 Luna model, which is less capable than Sol, and will receive unlimited text chats.
- According to OpenAI, factual error rates dropped by 62% for GPT-5.6 Luna and 68% for GPT-5.6 Sol compared to GPT-5.5 Instant, although these figures have not been independently verified.
Why it matters
The distinction between the models raises questions about accessibility and performance for free users, who will not have access to the more advanced reasoning capabilities of GPT-5.6 Sol. While the enhancements for paying users may improve their experience, the limitations imposed on free users could hinder their ability to effectively utilize AI for complex tasks. Furthermore, the claimed improvements in factual accuracy remain unverified, which may affect user trust in the models.
📖 Reader Mode
~2 min readOpenAI is giving Plus and Pro users a more focused GPT-5.6 Sol and a slider for adjusting response depth. Free users will get unlimited text chats with the cheaper GPT-5.6 Luna.
OpenAI says GPT-5.6 Sol now cuts unnecessary detail and excess formatting. For simple questions, it delivers a direct answer with just enough context, while more complex tasks get a fuller response that still makes the key recommendation easy to spot.
The company also claims better factual accuracy. In an internal evaluation using finance, medicine, and law prompts, responses with at least one factual error fell about 62 percent for GPT-5.6 Luna and 68 percent for GPT-5.6 Sol compared with GPT-5.5 Instant. Those figures haven't been independently verified.
The update only applies to ChatGPT. GPT-5.6 Sol remains unchanged in ChatGPT Work and Codex.
Paying users get more control over reasoning
A new slider lets paying users adjust how much effort ChatGPT puts into a response. Lower settings suit everyday questions, while higher settings are meant for planning, research, and coding. The slider was already available in ChatGPT Work.
OpenAI says it wants quick answers and deeper reasoning to feel like one model rather than two with different tones. An OpenAI developer recently broke down which of the five settings fits which type of task. The slider works on web, mobile, and desktop.
It's worth asking whether these controls actually help people use AI better, though. Users largely ignored the model switcher, and picking the right model for a task likely mattered way more than choosing among five reasoning levels. Adding more knobs to turn doesn't make the experience better.
Free users get more chats but a weaker model
GPT-5.6 Luna will become the default for Free and Go users later this week, according to OpenAI's release notes. Unlimited text chats and a new "Think" button for harder questions will follow next week, though limits on file uploads, image generation, and other tools will stay in place.
Luna may perform well for its size, but smaller models are generally more error-prone and less reliable. A case involving the previous Instant model, which produced a completely wrong data analysis compared to its larger reasoning counterpart, made that risk clear. Since the Think button only lets the smaller Luna reason longer rather than switching to a stronger model, free users lose access to OpenAI's most powerful reasoning altogether.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
