OpenAI Lifts ChatGPT Text Chat Limits for Free Users, Introduces Teen Safeguards

Post Views: 4

OpenAI has enhanced the GPT-5.6 Sol model, introducing improvements to response quality and user controls while expanding safeguards for younger audiences.

Enhancements to GPT-5.6 Sol Model

OpenAI has enhanced the GPT-5.6 Sol model, which powers ChatGPT for Plus and Pro users, introducing improvements to response quality and user controls. The update enables the model to handle both rapid responses and extended reasoning tasks without switching between distinct modes. For premium accounts, the GPT-5.6 Sol model now integrates instant and deep reasoning capabilities into a single framework. This change aims to produce more focused answers tailored to the complexity of user queries while minimizing unnecessary formatting. A new interface feature lets users adjust the depth of analysis applied to each response.

Free Tier Changes

Free users will no longer face text conversation rate limits, allowing uninterrupted interactions. However, restrictions on file uploads, image processing, and other tools remain in place under the company’s abuse prevention measures. Free tier users will transition to the GPT-5.6 Luna model by week’s end. Starting next week, these users will gain unlimited chat sessions with a dedicated “Think” button for complex queries.

Safeguards for Younger Audiences

OpenAI has expanded safeguards for users under 18, incorporating training to block content related to romantic roleplay, age-inappropriate challenges, and substitutes for human relationships. The model now includes boundaries around sexual content, eating disorder discussions, body image risks, and exposure to harmful activities. When detecting potential distress signals, the system is programmed to direct users toward appropriate support networks.

Security and Resilience Testing

Testing of GPT-5.6 Sol’s resilience against adversarial attacks revealed comparable performance to earlier versions. The model maintained a 1.000 score on connector evaluations, matching the GPT-5.5 model, while achieving 0.910 on search and function-calling tests. These assessments involved complex jailbreak scenarios and prompt injection attacks designed to override system instructions.

Ongoing Refinements

The updates follow internal evaluations focused on improving model behavior for younger demographics. OpenAI stated it continues refining responses through additional assessments and system-level protections. Internal testing demonstrated significant reductions in factual inaccuracies. When evaluated on financial, medical, and legal prompts, the model showed a 62% decrease in errors compared to the previous GPT-5.5 Instant version.