OpenAI has announced updates to ChatGPT that improve response quality, factual accuracy, and reasoning controls across its models. The update brings improvements to GPT-5.6 Sol for Plus and Pro users and expands GPT-5.6 Luna access for Free and Go users with new features.
GPT-5.6 Sol improvements
OpenAI said GPT-5.6 Sol has been updated for everyday ChatGPT usage, including quick questions, web searches, planning, research, advice, and complex decisions. The model is designed to provide more direct responses, use tighter formatting, and avoid unnecessary details when they do not add value.
The improvements include:
- Response length adjusted based on the complexity of the query.
- Shorter answers for simple questions with relevant context.
- More detailed responses for tasks such as planning, research, writing, coding, and decision-making while keeping the main recommendation clear.
- Corrections when needed instead of simply agreeing with incorrect assumptions.
Improved factual accuracy
OpenAI said GPT-5.6 Sol has been updated to reduce factual errors, especially for questions involving dates, numbers, sources, rules, and assumptions. The company said the model can better use the sources it finds while generating answers.
In an internal evaluation of financial, medical, and legal prompts requiring factual details, responses containing at least one factual error were about 68% less common with GPT-5.6 Sol and 62% less common with GPT-5.6 Luna compared with GPT-5.5 Instant.
Reasoning controls
OpenAI has brought ChatGPT’s Instant and Thinking experiences closer together with GPT-5.6 Sol. The company said moving from Instant responses to higher reasoning should provide more detailed answers while maintaining a consistent tone and behaviour.
Plus and Pro users can use a new reasoning slider in ChatGPT to control how much thought the model applies to each response. The slider can be adjusted based on the task:
- Lower reasoning levels for everyday questions.
- Higher reasoning levels for planning, research, writing, coding, and complex decisions.
The reasoning slider is available on ChatGPT web, mobile, and desktop.
GPT-5.6 Luna for free users
OpenAI is making GPT-5.6 Luna the default model for Free and Go users. The expanded access includes unlimited text chats and a new Think button that allows users to give the model more time to work through questions requiring deeper reasoning.
Safety updates
OpenAI has added safety training and evaluations focused on users under 18. The company said the model has been trained to avoid romantic roleplay, age-restricted challenges, and presenting itself as a substitute for real-world relationships.
Additional safety boundaries have been applied around:
- Sexual content.
- Eating disorders and body-image risks.
- Age-restricted goods.
- Dangerous activities.
- Graphic violence.
The model also encourages users to connect with trusted people when a teen may need support. OpenAI said these measures are supported by system-level protections and additional evaluations focused on younger users.
Availability
The updated GPT-5.6 Sol model and reasoning slider are available for Plus and Pro users in ChatGPT starting today.
GPT-5.6 Luna will roll out for Free and Go users with the following changes:
- GPT-5.6 Luna will become the default model for Free and Go users this week.
- Unlimited text chats and the new Think button will be available starting next week.
- Access to the Think button and unlimited text chats will remain subject to abuse safeguards.
- Limits will continue to apply for file uploads, images, and other tools.
OpenAI noted that this version of GPT-5.6 Sol is optimised for everyday ChatGPT conversations and will only be available within the Chat experience. The GPT-5.6 Sol version used for Work and Codex will remain unchanged as part of this release.