OpenAI Lifts ChatGPT Text Chat Limits for Free Users with Enhanced Teen Safeguards

Post Views: 1

OpenAI expands ChatGPT access for free users and enhances protections for younger demographics.

Expanded Access for Free Users

OpenAI has rolled out updates to its GPT-5.6 Sol model, which powers ChatGPT for Plus and Pro subscribers. The changes include modifications to text conversation limits for free-tier users and new safeguards targeting adolescent users. The company also introduced a mechanism to regulate the depth of reasoning applied during interactions. Free users will transition to the GPT-5.6 Luna model as their default setting this week. Starting next week, the platform will eliminate restrictions on text chat frequency, enabling continuous conversations without waiting for cooldown periods. A new “Think” button will be available to facilitate more complex reasoning tasks. While unlimited chats remain subject to abuse prevention measures, limitations persist for file uploads, image processing, and other advanced tools.

Transition to GPT-5.6 Luna

Free users will transition to the GPT-5.6 Luna model as their default setting this week.

New “Think” Button

A new “Think” button will be available to facilitate more complex reasoning tasks.

Limitations for Free Users

While unlimited chats remain subject to abuse prevention measures, limitations persist for file uploads, image processing, and other advanced tools.

Enhanced Safeguards for Younger Users

OpenAI has strengthened protections for users under 18, incorporating training to avoid content related to romantic roleplay, age-inappropriate challenges, and substitute relationships. The model now includes boundaries to address sexual material, eating disorder discussions, body image risks, dangerous activities, and graphic violence. When detecting potential distress, the system prioritizes directing users toward trusted human connections.

Content Restrictions for Minors

Training to avoid content related to romantic roleplay, age-inappropriate challenges, and substitute relationships.

Distress Detection and Support

When detecting potential distress, the system prioritizes directing users toward trusted human connections.

Model Improvements and Security Evaluations

For Plus and Pro accounts, GPT-5.6 Sol now integrates both rapid response and extended analytical capabilities within a single model. This eliminates the need to switch between distinct modes for instant replies and in-depth reasoning. The updated system aims to produce more focused responses, adjust detail levels based on query complexity, and minimize unnecessary formatting. A configurable slider allows users to specify the extent of reasoning effort applied to each answer. Internal testing demonstrated improved accuracy with the new model. Evaluations on financial, medical, and legal prompts revealed a 62% reduction in factual errors with GPT-5.6 Luna and a 68% decrease with GPT-5.6 Sol compared to the previous GPT-5.5 Instant model. Security evaluations focused on resilience against adversarial attacks. GPT-5.6 Sol was tested against jailbreak attempts designed to bypass refusal protocols, with results comparable to earlier versions. The model maintained a 1.000 score on connector evaluations, matching GPT-5.5 and surpassing the 0.998 score of GPT-5.4 Thinking. On search and function-calling resistance tests, GPT-5.6 Sol achieved a 0.910 rating.

Configurable Reasoning Slider

A configurable slider allows users to specify the extent of reasoning effort applied to each answer.

Internal Testing Results

Evaluations on financial, medical, and legal prompts revealed a 62% reduction in factual errors with GPT-5.6 Luna and a 68% decrease with GPT-5.6 Sol compared to the previous GPT-5.5 Instant model.

Security Evaluations

Security evaluations focused on resilience against adversarial attacks. GPT-5.6 Sol achieved a 0.910 rating on search and function-calling resistance tests.

Balancing Accessibility and Safety

The updates reflect ongoing efforts to balance accessibility with safety, particularly for younger audiences. OpenAI emphasized continued refinement of model behavior through system-level safeguards and targeted assessments for users under 18.