OpenAI Gives Free Users Unlimited Chats and a Smarter Model

OpenAI announced on August 6, 2026, that GPT-5.6 Luna will become the default model for Free and Go users, replacing the older GPT-5.5 Instant. Starting the week of August 10, those users will also get unlimited text chats and access to a new Think button that triggers higher-effort reasoning on a per-message basis. Paid subscribers on Plus and Pro plans receive an updated version of GPT-5.6 Sol alongside a new slider that lets them dial in exactly how much reasoning the model applies to any given response.

The changes are the most significant structural upgrade to ChatGPT’s free tier in some time. Before this announcement, free users faced a cap of roughly 10 messages per five-hour window — a hard stop that made iterative workflows like debugging code, revising essays, or working through a multi-step problem set genuinely painful. That cap is now gone for text conversations, though limits remain in place for file uploads, image analysis, voice and other tools. OpenAI also noted the unlimited access is subject to abuse guardrails.

What’s Actually New for Free Users

GPT-5.6 Luna is a step up in model quality over what free users previously had access to by default. In an internal evaluation of financial, medical and legal prompts requiring factual precision, responses with at least one factual error were about 62% less common with Luna than with GPT-5.5 Instant. That caveat — internal evaluation, not independent testing — is worth holding onto, but the directional improvement is meaningful for anyone using ChatGPT to research health questions, understand legal language, or work through quantitative problems.

The Think button adds another layer. Rather than routing all queries through the same process, free users can now flag specific questions as ones that deserve more careful reasoning. That on-demand access to higher-effort thinking without a paid subscription is new territory for OpenAI’s free offering.

What’s Changing for Plus and Pro Subscribers

The updated GPT-5.6 Sol model for paid users is tuned specifically for the conversational use case inside ChatGPT’s Chat interface — not the versions of Sol powering Work or Codex, which are unchanged. The main improvements are tighter, more direct answers and better factual reliability. In the same internal evaluation framework, Sol reduced error-containing responses by roughly 68% compared to GPT-5.5 Instant.

The new effort slider replaces what were previously two distinct modes — Instant and Thinking — with a single adjustable experience. OpenAI’s intent is that moving from a quick answer to a more considered one should feel like the same model taking more time, rather than a jarring switch between two different personalities or styles.

The announcement also details expanded teen safety protections. Model-level training now steers users believed to be under 18 away from romantic roleplay, age-restricted challenges, and interactions that position the AI as a substitute for real-world relationships. OpenAI says it has also added new evaluation benchmarks specifically measuring how its models perform for younger users.

Where This Lands Competitively

All four major AI assistants — ChatGPT, Claude, Copilot and Gemini — maintain permanent free tiers. Prior to this announcement, ChatGPT’s message cap was its most visible structural disadvantage relative to Google’s Gemini and Microsoft’s Copilot, both of which are essentially unlimited on the free tier (though throttled under heavy use). Claude imposes daily message limits on its free tier. By removing the text cap and simultaneously pushing a newer-generation model to free users, OpenAI is making an access argument that none of its main rivals have yet matched at equivalent model quality. For long-form writing and analysis, Claude’s free tier still holds a strong reputation, and Gemini 3.1 Pro offers a 1 million token context window at the paid level. But on the raw question of how much text you can generate for free with a capable model, OpenAI has moved ahead.

What This Means If You’re a Student

The removal of the text rate limit is the most immediately practical change. Students who were rationing messages across study sessions, toggling between free accounts, or paying $20 a month primarily to avoid the cap now have a clearer calculation to make. Unlimited text chats with GPT-5.6 Luna covers the core iterative workflows — writing and revising drafts, stepping through code problems, building out research outlines — without requiring a subscription.

The Think button extends that calculus for harder questions. Need to work through a pharmacology concept, parse a contract clause, or check the logic in a statistics problem? Tapping Think gives the model more time to reason through the answer rather than rushing to a response. The ~62% reduction in factual errors is a relevant figure here, though it’s still worth verifying anything high-stakes against a primary source.

The meaningful caveat: if your workflow depends on uploading PDFs, analyzing images, or using voice features regularly, the free tier still has constraints. Those tool-based limits haven’t moved. For document-heavy research or multimodal tasks, Plus remains the more capable option. But for text-first students who’ve been on the fence, the new free tier is a considerably stronger starting point than it was a week ago.

Source: OpenAI