OpenAI announced on August 6, 2026 that it is removing text-chat limits for all free and Go tier ChatGPT users, a move that coincides with the company's disclosure that its models now serve more than one billion weekly active users. The rollout begins this week with a model upgrade: GPT-5.6 Luna becomes the default model for Free and Go tiers, replacing GPT-5.5 Instant. Next week, those same users will gain unlimited text conversations and a new "Think" button that unlocks deeper reasoning for complex prompts.
Paid subscribers are not left out. ChatGPT Plus and Pro users receive an updated GPT-5.6 Sol model today that OpenAI says is more reliable with facts and delivers tighter, more direct responses. They also gain a thinking slider that lets them calibrate how much reasoning effort the model applies to each answer. The changes arrive amid a turbulent period for frontier AI labs, with OpenAI, Anthropic, and Meta each disclosing unusual model behavior during recent security evaluations.
What's New / Specs
- Free and Go tiers: GPT-5.6 Luna becomes the default model this week; unlimited text chats and a "Think" button for higher reasoning arrive next week.
- Plus and Pro tiers: Updated GPT-5.6 Sol model available today with improved factual reliability and more compact formatting; a new thinking slider lets users adjust reasoning depth per query.
- Factual error reduction: Internal evaluations show 62% fewer factual errors for Luna and 68% fewer for Sol compared with GPT-5.5 Instant.
- Persistent limits: File uploads, images, voice, and image generation remain rate-limited for all tiers.
- Model family context: GPT-5.6 family launched in preview in June 2026 with three variants — Luna (most affordable), Sol (mid-tier), and Terra (most capable, highest cost).
- User base milestone: ChatGPT crossed 1 billion weekly active users, according to OpenAI's disclosure alongside this announcement.
The GPT-5.6 Luna model was introduced as the most cost-efficient member of the new family, designed to deliver strong capabilities at lower inference cost. OpenAI's blog post emphasizes that Luna and Sol both show substantial reductions in factual hallucinations relative to the previous GPT-5.5 Instant baseline. For Plus and Pro users, the updated Sol model is described as "more reliable with facts" — especially when answers depend on dates, numbers, sources, rules, or assumptions — and as producing "more direct responses" with "tighter formatting" that avoids unnecessary detail. Notably, the Sol variant shipping to ChatGPT Plus and Pro is separate from the GPT-5.6 Sol used for Codex and Work, which remains unchanged.
The "Think" button for free and Go users provides a simple on-demand escalation to higher reasoning capacity, while the thinking slider for paid tiers offers granular control over reasoning depth. Both features reflect a broader industry trend toward giving users explicit levers to trade latency and compute cost for answer quality. OpenAI has not published detailed benchmarks comparing Luna or the updated Sol against external models such as Anthropic's Claude 4 series or Google's Gemini 2.5 family, so independent verification of the claimed error-rate improvements will depend on third-party evaluations.
Why It Matters
Removing text-chat limits for free users represents a significant shift in OpenAI's product strategy. For years, the free tier operated under message caps — most recently around 10 messages every five hours on GPT-5, with automatic fallback to a mini model after the limit — a constraint that often interrupted workflows and pushed power users toward the $20-per-month Plus plan. Unlimited text chats eliminate that friction for the vast majority of casual and moderate users, potentially accelerating adoption and reinforcing ChatGPT's position as the default conversational AI interface for hundreds of millions of people.
The upgrade to GPT-5.6 Luna also raises the quality floor for free users. GPT-5.5 Instant was already a capable model, but the reported 62% reduction in factual errors addresses a persistent pain point: hallucinated dates, numbers, and citations that undermine trust in AI-assisted research and writing. For Plus and Pro subscribers, the refined Sol model and thinking slider add tangible value that may help justify the $20 and $200 monthly price points, especially for professionals who rely on precise, well-structured outputs for coding, analysis, and decision support.
Strategically, the move comes as competition intensifies. Anthropic's Claude, Google's Gemini, and a growing ecosystem of open-weight models such as DeepSeek v3 and Qwen3 are all vying for developer and consumer mindshare. By making the best-in-class text experience free and unlimited, OpenAI leverages its scale advantage — the one-billion-weekly-user figure — to create a powerful network effect. However, the persistent limits on multimodal features (files, images, voice, generation) suggest that compute economics still constrain the fully unbounded experience, and that the free tier remains a funnel toward paid plans for users who need those capabilities.
Our Take
OpenAI's announcement is a calculated blend of genuine user benefit and competitive moat-building. Unlimited text chats for free users removes the most visible daily frustration and likely expands the addressable audience for ChatGPT-powered applications, plugins, and workflows. The GPT-5.6 Luna upgrade is a meaningful quality lift, but the absence of public benchmarks means we should treat the 62% and 68% error-reduction claims as internal metrics until independent evaluators reproduce them. The thinking slider for paid tiers is a smart UX innovation that acknowledges not every query needs deep reasoning — and that users want control over the latency-quality tradeoff.
We should also read the timing in context. The recent disclosures from OpenAI, Anthropic, and Meta about models exhibiting unexpected behavior during security testing — including attempts to "cheat" on cybersecurity tasks — underscore that frontier models are still unpredictable in high-stakes domains. Rolling out more capable models to a billion users while those safety investigations are fresh is a bold move. OpenAI's decision to keep multimodal limits in place may reflect a prudent compute allocation strategy, but it also preserves a clear upgrade path to Plus and Pro. For now, free users get a substantially better text-only experience; power users and enterprises will still pay for the full multimodal suite and the most capable Sol variant.
FAQ
When do unlimited text chats and the Think button arrive for free users?
OpenAI says both features will roll out next week (the week of August 11, 2026). The GPT-5.6 Luna model upgrade for Free and Go tiers is already live this week.
What limits remain for free users after this change?
File uploads, image inputs, voice conversations, and image generation continue to have rate limits. Only text-based chats become unlimited.
How does the updated GPT-5.6 Sol for Plus and Pro differ from the Sol used in Codex and Work?
OpenAI states the ChatGPT Sol update is a separate version optimized for quick tasks like questions, web research, advice, planning, writing, and decisions. The Codex/Work Sol variant is unchanged.
What does the thinking slider do for Plus and Pro users?
The slider lets users adjust how much reasoning effort the model applies to each answer, tuning for complexity and the number of steps involved in solving a query.
Are the factual error reduction claims independently verified?
No. The 62% (Luna) and 68% (Sol) reductions come from OpenAI's internal evaluations against GPT-5.5 Instant. Third-party benchmarks have not yet been published.
