OpenAI Ends Free ChatGPT Text Limits With Luna Model

OpenAI is removing text rate limits for Free and Go ChatGPT users, defaulting them to GPT-5.6 Luna and adding a Think button for harder questions, the company said on August 6. The move lands as ChatGPT crosses 1 billion weekly users and arrives alongside a tighter GPT-5.6 Sol update and reasoning slider for Plus and Pro.

Limits stay in place for files, images, voice and image generation. The free text faucet is the real change.

What Free and Go Users Receive This Week

GPT-5.6 Luna becomes the default model for Free and Go users this week. Next week those same accounts get unlimited text chats and the new Think button, subject to abuse guardrails.

OpenAI’s August 6 product post frames the shift as expanding access while improving everyday usefulness. Luna is the fast, low-cost member of the GPT-5.6 family. Free users previously hit dynamic Instant caps that could reset after a few hours; plain text no longer meters that way.

Think gives Luna extra time on complex prompts. It is not full Sol access. File uploads, images and other tools keep their ceilings. The split is deliberate: open-ended conversation becomes abundant, while compute-heavy side channels stay metered.

• Unlimited: everyday text chats with GPT-5.6 Luna
• Still limited: file uploads, image generation, voice, deep research and certain tool calls
• New control: Think button for harder questions (starts next week)
• Guardrails: abuse protections can still throttle or restrict accounts

The current ChatGPT Free and Go tiers already list unlimited text chats with Luna on the Free plan. Go at $8 a month still sells more tool messages, uploads, image creation, voice and longer memory, and may include ads.

For casual users the practical effect is simple. A long brainstorming thread, a homework explanation, or a multi-turn planning chat no longer risks a soft lock after a burst of messages. The account still meets walls when the prompt attaches a file or calls a tool, so the free experience stays text-first by design.

Fewer Factual Errors on High-Stakes Prompts

OpenAI ran an internal evaluation on financial, medical and legal prompts that demand factual detail. Responses containing at least one factual error were about 62% less common with GPT-5.6 Luna and 68% less common with GPT-5.6 Sol than with GPT-5.5 Instant.

Reliability snapshot

• 62% fewer error-containing responses for Luna vs GPT-5.5 Instant
• 68% fewer for the updated Sol
• Eval domains: finance, medicine, law
• Claim is internal; external benchmarks are not yet public in the announcement

Sol also aims for tighter formatting and answers that lead with the real question. OpenAI’s own bike-commute example shows Sol naming wind as the main issue and skipping the padded forecast lists that Instant produced.

That consistency matters because the same Sol model now powers Instant-style and deeper replies for paid users, so tone and behavior stay closer when someone raises the effort level.

Luna’s 62% figure still matters for the free tier. Everyday users inherit a default that OpenAI’s own high-stakes set already scores well below the prior Instant baseline, even before anyone presses Think. Sol’s further cut to 68% is the paid differentiator on the same internal yardstick, not a claim that free answers are unreliable.

Plus and Pro Get Sol Everywhere Plus a Slider

Plus and Pro users already have the updated GPT-5.6 Sol and the new thinking slider on web, mobile and desktop. The slider lets them dial how much thought goes into each answer: keep it quick for simple queries or push it up for planning, research, writing or multi-step decisions.

This Chat-optimized Sol is separate from the Sol used in Work and Codex, which is unchanged. OpenAI says the Chat version is tuned for focused everyday answers and will not appear outside the main chat experience.

According to GPT-5.6 availability by plan, Free and Go still cannot select Sol in standard conversations. Their Think path stays on Luna. Plus gets Medium and High reasoning; Pro adds Extra High and Sol Pro for the hardest work.

Prices and tool quotas still differ sharply. The text meter is no longer the main free-vs-paid wall.

The slider’s value is continuity. A Plus user can stay inside one Sol-powered thread, nudge effort up for a dense research turn, then drop back for quick follow-ups without switching models or losing tone. Free users get a binary Think boost on Luna instead of that continuous control.

The Meter That Still Clicks

OpenAI is clear that files, images, voice and image generation keep separate limits. Free users also face tighter context windows, slower or limited image gen, limited memory and restricted Codex and Work access compared with paid tiers.

That is where conversion pressure now lives. Casual Q&A and brainstorming can run all day on Luna. Anyone who needs heavy uploads, high-volume images, long memory or maximum reasoning still hits a ceiling that Go, Plus or Pro raise.

• Free: unlimited Luna text; tighter context, limited memory, capped media and tools
• Go ($8): same Luna text floor, plus more tool messages, uploads, images, voice and longer memory; ads possible
• Plus ($20): Sol default, thinking slider, projects, custom GPTs and early features
• Pro (from $100): multiplied usage, full slider range and Sol Pro

Go sits in the middle: expanded tool and media headroom at $8, with the possibility of ads. Plus keeps the advanced Sol path, projects, custom GPTs and early features. Pro multiplies usage and unlocks Sol Pro.

Text abundance does not erase those walls. It only moves the first “upgrade?” moment later in the session, to the instant someone attaches a PDF, asks for a batch of images, or wants Sol-grade depth on every turn.

Why the Free Floor Just Moved for Everyone

Removing the text ration does more than delight free users. It turns ChatGPT into a reliable daily default for ordinary conversation at a moment when 1 billion people already open it each week. Habit formation gets cheaper for OpenAI and stickier for everyone else to disrupt.

Rivals that still throttle free text or keep weaker defaults now look stingy by comparison. The competitive fight shifts from “can I keep chatting?” to “is the quality gap big enough to notice and pay for?” That same pressure shows up in how people already compare systems when they how the big models rate one another.

On X, the loudest practical request was not more free chat. It was the same unlimited Luna treatment for Codex. One widely liked reply offered to drop Claude immediately if coding quotas matched the new free text faucet. The crowd is already treating open-ended text as table stakes and hunting the next scarce resource.

This is a concrete step toward more abundant intelligence: making our latest models more widely available, improving the usefulness and reliability of the answers people get, and letting free users keep text chats going without a rate limit.

OpenAI wrote that line in the announcement. Access shapes opportunity, the company added. The second-order bet is that abundance on the free side grows the overall pie faster than it cannibalizes Plus, especially while tools, images and controllable depth stay gated.

Regulators are watching distribution too. Separate EU pressure to open Android AI access already forces platform owners to leave room for rivals. An unlimited free ChatGPT raises the bar those rivals must clear on mobile and desktop alike.

How Think and the Slider Split Effort

Both free and paid paths now expose a way to spend more compute on a hard prompt. The mechanisms differ in grain and in which model runs underneath.

Free and Go get a Think button on Luna. It is an on-or-off boost for complex questions, not a continuous dial, and it never promotes the thread to Sol. Plus and Pro get the thinking slider on Sol, with Medium and High on Plus and Extra High plus Sol Pro on Pro.

That design keeps the free tier simple while reserving fine-grained control for subscribers. A student who hits a tough step can tap Think and stay on Luna. A professional who lives in long planning threads can ride the slider without leaving Sol’s formatting and error profile.

Work and Codex still use their own unchanged Sol path. The Chat-tuned Sol and its slider stay inside the main chat experience, so agentic and coding surfaces do not automatically inherit the new everyday defaults.

What the Quality Gap Must Justify Now

Once unlimited Luna text ships, paid plans can no longer lean on “you ran out of messages” as the primary hook. They have to win on dimensions the free tier still rations.

Those dimensions are already listed in the product split: Sol’s higher internal factual-error cut, continuous reasoning control, wider context and memory, faster or larger image and voice quotas, projects, custom GPTs, Codex and Work access, and Sol Pro at the top. The pricing table still runs from $0 to from $100; only the meaning of the bottom rung changed.

If free users rarely notice the Sol gap on ordinary tasks, conversion may slow. If they hit tool ceilings or want slider-level depth several times a week, Go and Plus remain the relief valves. The next quarter’s retention and upgrade numbers will show which pattern dominates after the text meter disappears.

Rollout Calendar and What Stays the Same

1. August 6, 2026: Updated GPT-5.6 Sol and thinking slider live for Plus and Pro in Chat.
2. This week: GPT-5.6 Luna becomes default for Free and Go.
3. Next week: Unlimited text chats and Think button reach Free and Go (abuse guardrails apply).
4. Unchanged: Sol version powering Work and Codex; separate limits on files, images, voice and generation; under-18 safety training and system-level protections described in the accompanying system card.

Safety notes in the post highlight extra training against romantic roleplay with minors, age-restricted challenges, and presenting the model as a substitute for real relationships, plus boundaries around other high-risk topics. Those apply across the new defaults.

The product change is already visible on the pricing page. Free no longer sells itself primarily as a capped trial of Instant. It sells unlimited text with a modern lightweight model and a deeper-reasoning button for when the query needs it.

Paid tiers still buy speed, capacity, Sol-grade consistency, the slider, and the multimodal and agentic features that text alone cannot replace. Whether that package stays compelling once free text never ends is the open commercial question the next quarter will answer.

Frequently Asked Questions

What does unlimited text chats mean on free ChatGPT?

It means Free and Go accounts no longer face rate limits on ordinary text messages with GPT-5.6 Luna once the feature rolls out next week. Messages that include files, images or other tools still count against separate quotas, and abuse guardrails can still restrict accounts that violate the terms.

How is GPT-5.6 Luna different from GPT-5.6 Sol?

Luna is the fastest and lowest-cost model in the GPT-5.6 family and becomes the free default. Sol is the higher-capability model reserved for Plus, Pro and higher plans in standard chat; OpenAI reports it cut factual-error responses by 68% versus GPT-5.5 Instant on its internal high-stakes set, compared with 62% for Luna.

Do free users get the thinking slider?

No. Free and Go users receive a Think button that spends more compute on harder questions using Luna. The continuous thinking slider that adjusts effort level is for Plus and Pro users running Sol.

How much does ChatGPT Go cost and what does it add?

Go is $8 per month in the US and includes everything in Free plus higher limits on tool messages, uploads, image creation, voice chats and longer memory. The plan may include ads. It still uses Luna for standard chat rather than Sol.

When do the free unlimited chats and Think button arrive?

Luna becomes the Free and Go default this week. Unlimited text chats and the Think button are scheduled for next week, according to OpenAI’s August 6 announcement, with the exact day varying by account as the rollout completes.