OpenAI has announced a significant update for its ChatGPT platform, aimed at enhancing user experience and expanding access for its one billion weekly users. The changes introduce new, more powerful models for both free and paid tiers, alongside features designed to improve conversational quality and factual reliability. This move eliminates text-based chat limits for free users, marking a major step in making advanced AI more widely available.
Enhanced Experience for All Users
Free and Go users will now default to the new GPT-5.6 Luna model, a direct replacement for the previous GPT-5.5. This upgrade is complemented by the removal of limits on text-based conversations, allowing for unrestricted interaction. For more demanding queries, a new "Think" button provides access to higher reasoning capabilities, giving the model more time to formulate a comprehensive answer.
Advanced Features for Subscribers
Subscribers to ChatGPT Plus and Pro are also receiving substantial upgrades, gaining access to an improved GPT-5.6 Sol model. This version is engineered to deliver more compact, robust, and factually reliable answers for a range of quick tasks like planning and research. The update aims to create a more consistent and efficient experience for users who rely on the platform for professional and complex work.
A key new feature for paid subscribers is the introduction of a "thinking slider," which offers granular control over the model's computational effort. Users can adjust the slider based on the complexity of their query, requesting quicker responses for simple questions or deeper analysis for multi-step problems. This tool allows for a more tailored interaction, optimizing for either speed or depth depending on the user's immediate needs.
A Leap in Factual Accuracy
A core focus of this update is the significant improvement in factual accuracy across the new models. According to OpenAI's internal evaluations, factual errors were found to be 62% less common with GPT-5.6 Luna compared to its predecessor. The upgraded GPT-5.6 Sol for subscribers demonstrated an even greater improvement, with a 68% reduction in factual inaccuracies.
This enhancement in reliability is crucial for tasks involving specific data such as dates, numbers, and sources. The models are now better designed to utilize the information they find to construct more dependable answers. This push for greater accuracy addresses a key challenge in AI development and aims to build greater user trust in the platform's outputs.
Rollout and Availability
The rollout of these new features is being staggered over the coming weeks. Plus and Pro users can access the updated GPT-5.6 Sol and the thinking slider immediately. The new GPT-5.6 Luna model will become the default for Free and Go users this week, with unlimited text chats and the "Think" button following next week.
OpenAI has clarified that while text chat limits are being removed for free users, restrictions will remain in place for other functionalities. This includes limits on file uploads, image analysis, and image generation tools. The company also noted that the version of GPT-5.6 Sol used for its specialized Work and Codex products remains unchanged by this update.
This comprehensive update from OpenAI represents a strategic move to both broaden the accessibility and deepen the capability of ChatGPT. By removing barriers for free users and providing more powerful tools for subscribers, the company reinforces its commitment to its mission of democratizing AI. These enhancements in performance, accuracy, and user control are set to redefine the standard for conversational AI for its vast global user base.