OpenAI is eliminating text-only conversation caps for Free and Go ChatGPT users while introducing updated reasoning models and interface tools across its platform. These changes include the rollout of GPT-5.6 Luna for entry-level tiers and an enhanced GPT-5.6 Sol model for paid subscribers.
Expanded Access to Free Tiers
OpenAI has officially announced the removal of message frequency caps for standard text-based interactions on its Free and Go tiers. This development comes as the platform reports reaching a milestone of 1 billion weekly active users. While users will no longer face rate limits for basic text chats, OpenAI clarified that resource-intensive interactions—such as voice mode, file uploads, image generation, and image analysis—will still remain subject to specific usage constraints. This strategic shift is designed to lower the barrier for casual users who rely on the service for high-volume text conversations, allowing for a more fluid experience without the fear of hitting a sudden service ceiling.
Technical Model Upgrades
A significant portion of the announcement involves the transition to new underlying architecture. Users on the Free and Go plans are seeing their default experience move from the legacy GPT-5.5 model to the more advanced GPT-5.6 Luna. Simultaneously, Plus and Pro subscribers gain access to an updated iteration of the GPT-5.6 Sol model. Internal testing conducted by the company suggests these upgrades represent a substantial improvement in accuracy. Specifically, data shows that factual error rates have dropped by 62% when utilizing the Luna architecture and by 68% when using the refined Sol model compared to the previous GPT-5.5-Instant benchmarks. The new Sol version is explicitly tuned to handle complex logic, such as rules, dates, and citation synthesis, with greater precision.
New Reasoning Capabilities
To complement the model updates, OpenAI is introducing new interface tools that give users finer control over how the AI processes information. Free and Go users will gain access to a 'Think' button starting next week. This feature allows users to trigger a deeper reasoning process when they encounter particularly complex inquiries. Meanwhile, paying subscribers on Plus and Pro tiers are receiving a more sophisticated 'thinking slider.' This tool empowers users to manually adjust the intensity and duration of the model's 'thought process' before it generates a final output. By allowing users to toggle this setting based on the complexity of their query, the company aims to balance the need for rapid responses with the requirement for meticulous, step-by-step logical analysis.
Deployment Timeline and User Experience
The rollout of these features is staggered throughout the week of August 6, 2026. Certain improvements for Plus and Pro users—specifically the updated GPT-5.6 Sol model and the new thinking slider—are scheduled for immediate availability as of Thursday. For Free and Go tier users, the transition to the GPT-5.6 Luna model begins this week, while the activation of unlimited text messaging and the 'Think' reasoning button will follow during the subsequent week. The company notes that the version of GPT-5.6 Sol being deployed to ChatGPT users is distinct from the specialized enterprise-grade versions used for Codex and professional workspace tools, ensuring that standard chat performance remains distinct from business-specific infrastructure requirements.
⚖ The Balanced View
Supporting view
The update directly addresses common user friction by removing text limits and provides tangible accuracy improvements of over 60% according to internal metrics.
Concerns & criticism
Despite the removal of text limits, users on the Free and Go tiers will still encounter usage restrictions when interacting with media-heavy features like image generation or file uploads.
→What's next
Users on the Free and Go tiers should look for the 'Think' button to appear in their interface early next week. Meanwhile, those subscribed to the Plus and Pro services can immediately experiment with the thinking slider to optimize their query outputs for accuracy and complexity.