What actually changed
If you've been waiting for ChatGPT Free to stop cutting you off mid-session, this is the update that matters. OpenAI is making GPT-5.6 Luna the default for Free and Go accounts, and it's removing the usual rate limit on normal text chats for eligible users — with abuse guardrails still in place.
There's a second half for paid users. GPT-5.6 Sol inside ChatGPT Chat is getting a more direct writing style, and Plus / Pro can dial reasoning effort with a slider. Same family of models, different controls depending on your plan.
Read "unlimited" carefully. OpenAI is expanding everyday text conversations. File uploads, image tools, and other features still keep their own caps. Spam and abuse can still get restricted.
ChatGPT August 2026 update at a glance
| Area | What's changing | Who gets it | Catch |
|---|---|---|---|
| Default model | GPT-5.6 Luna becomes the standard | Free and Go | Phased rollout — not every account at once |
| Text chats | Unlimited normal text conversations | Free and Go | Abuse guardrails still apply |
| Deeper reasoning | Think button for harder prompts | Free and Go | Tool / media limits stay separate |
| Paid chat model | Updated GPT-5.6 Sol — tighter, more direct answers | Plus and Pro | This drop is for ChatGPT Chat |
| Reasoning control | Slider for how much thought a reply gets | Plus and Pro | May appear gradually during rollout |
What GPT-5.6 Luna means for Free users
Luna is the fast, efficient model in the GPT-5.6 lineup. As the Free / Go default, it's meant for the stuff people actually do all day: studying, rewriting drafts, language practice, planning, and a long chain of follow-ups where you don't want the conversation to die because you hit a message wall.
It doesn't mean every prompt gets the same compute. Quick questions stay quick. When you need more depth, you tap Think instead of hunting through a model picker.
That's the design bet: keep casual chat lightweight, and make deeper reasoning an obvious switch rather than a maze of model names.
How the Think button changes the experience
- Everyday mode — explanations, rewrites, summaries, and straightforward questions without waiting around.
- Think mode — extra reasoning time for multi-step planning, comparisons, and tougher problem-solving.
- Less model jargon — you don't need to memorize which GPT-5.6 variant is "smarter" for a hard prompt.
- Still guarded — higher reasoning access sits behind the same product safeguards as the rest of ChatGPT.
If you're tutoring yourself through a topic or debugging a long plan, Think is the button worth using. For "rewrite this email nicer," everyday mode is usually enough.
What's new for ChatGPT Plus and Pro
Plus and Pro users get an updated GPT-5.6 Sol inside ChatGPT Chat. OpenAI's pitch is straightforward: answer the main question earlier, keep formatting tighter, and cut the filler. The same model is supposed to handle both quick replies and deeper reasoning, so switching modes shouldn't feel like switching products.
The new reasoning slider is the paid control that Free doesn't get. Lower settings favor speed for routine work. Higher settings buy more analysis for research, coding, writing, and decisions with real trade-offs.
| Task | Suggested setting | Why |
|---|---|---|
| Quick factual question | Standard / lower | Faster reply, enough context |
| Long-form planning | Higher | Room for dependencies and sequencing |
| Research synthesis | Higher (+ check sources) | Better handling of dates, claims, and assumptions |
| Coding or debugging | Higher | Multi-step inspection and validation |
| Simple rewrite | Standard | Extra depth mostly adds latency |
One clarification that trips people up: OpenAI says the GPT-5.6 Sol version powering Work and Codex is not changing with this ChatGPT Chat release. Don't assume every OpenAI product flipped overnight.
OpenAI's factual reliability claims
OpenAI reports an internal evaluation on financial, medical, and legal prompts that need factual detail. Compared with GPT-5.5 Instant, responses with at least one factual error were less common on the new models. Those are company-reported numbers under OpenAI's test conditions — not a promise that your next answer will be perfect.
| Model | Reported change vs GPT-5.5 Instant | How to read it |
|---|---|---|
| GPT-5.6 Luna | ~62% fewer responses with ≥1 factual error | Internal eval result, not a universal accuracy score |
| GPT-5.6 Sol | ~68% fewer responses with ≥1 factual error | Internal eval under OpenAI's conditions |
For medical, legal, or money decisions, verify elsewhere. Better eval numbers lower risk. They don't delete hallucinations, stale facts, or bad assumptions.
What unlimited text chats do not include
- File uploads are still limited.
- Image generation and image analysis keep separate quotas.
- Other tools can have their own caps, availability rules, or regional limits.
- Automated abuse, spam, scraping, and policy violations can still get blocked.
- This is not unlimited API usage — ChatGPT plans and API billing are different products.
Rollout timeline and availability
| Feature | Announced availability | Note |
|---|---|---|
| Updated GPT-5.6 Sol | Started August 6, 2026 | Plus and Pro in ChatGPT |
| Reasoning slider | Started August 6, 2026 | Web, mobile, and desktop for eligible paid users |
| GPT-5.6 Luna as default | Week of August 6 | Free and Go |
| Unlimited text + Think button | Beginning the following week | Phased deploy + abuse guardrails |
If your account still looks unchanged, wait a few days before assuming you were skipped. These rollouts usually land in waves.
Why this update matters
The practical win is continuity. A student can keep asking follow-ups. A writer can iterate a draft twenty times. A small business can stay in one planning thread without watching a message counter.
For OpenAI, the move lowers friction on text — the cheapest, stickiest habit — while still metering the expensive stuff: files, images, and tool-heavy work.
It also simplifies the product story. Free users get Luna by default and an obvious Think option. Paid users get Sol plus finer control through the slider. Less time picking models. More time actually using the chat.
Frequently asked questions
Are ChatGPT Free messages now completely unlimited?
Normal text chats are expanding without the usual rate limit, but abuse guardrails still apply. Images, files, and other tools keep separate limits.
What is GPT-5.6 Luna?
GPT-5.6 Luna is OpenAI's fast, efficient model in the GPT-5.6 family. It's becoming the default for ChatGPT Free and Go users.
What does the Think button do?
It gives GPT-5.6 Luna more reasoning time when a question needs deeper analysis — planning, comparisons, multi-step problem solving.
Does this update change Codex or ChatGPT Work?
OpenAI says the GPT-5.6 Sol version used by Work and Codex is not changing as part of this specific ChatGPT Chat release.
Do Plus and Pro get unlimited text chats too?
This announcement focuses unlimited text chats on Free and Go. Plus and Pro get the Sol quality update and the reasoning slider instead.
Bottom line: Free and Go get a lot more room for everyday text. Plus and Pro get a cleaner Sol experience and control over reasoning effort. Text access is expanding in a big way — files, images, tools, and misuse protections are still limited on their own tracks.
Also see our ChatGPT app page on AiSulivo for listing details, and confirm live limits inside your own ChatGPT account.
Back to top