ChatGPT Drops Text Limits on Free Accounts and Makes GPT-5.6 Luna the Default

OpenAI is removing the cap on how many text messages free ChatGPT accounts can send, and switching those accounts to a newer default model called GPT-5.6 Luna. The company announced both changes on Thursday, alongside an updated GPT-5.6 Sol for Plus and Pro subscribers and a new reasoning slider that replaces the old split between Instant and Thinking modes.

The two free-tier changes do not land at the same moment, which is the detail most likely to confuse people this week. Luna becomes the default model for Free and Go accounts during this week; the removal of text chat limits is scheduled to begin the week after.

What "unlimited" actually covers

The change applies to plain text conversations only, and OpenAI attaches an abuse-guardrail caveat to it, so this is not a licence to run automated traffic through a free account. Everything that costs more to serve stays metered.

• Unlimited: text-only chats on Free and Go accounts, plus the new Think button, which hands a harder question more processing time before answering.
• Still limited: file uploads, image uploads, image generation, voice chats and other tools.

In practice that means a free account can hold a long written conversation without hitting a wall, but attaching a PDF or generating a picture will still run into a quota. The Go tier, OpenAI's cheaper paid plan, is covered by the same change as the free tier.

The model swap underneath

GPT-5.6 Luna replaces GPT-5.5 Instant as the default for Free and Go. OpenAI's own internal evaluation, run on financial, medical and legal prompts that require precise factual detail, put factual errors 62 percent lower with Luna than with GPT-5.5 Instant. Those are the company's figures on its own test set rather than an independent benchmark, so treat them as a direction of travel rather than a guarantee.

Paid subscribers get a separate update. The refreshed GPT-5.6 Sol is tuned for everyday conversation — more direct answers, tighter formatting, less padding when extra detail does not help — and OpenAI reports factual errors 68 percent lower than GPT-5.5 Instant on the same prompt set.

A slider instead of two modes

Plus and Pro users also get a reasoning slider, available on the web, mobile and desktop apps. Instead of choosing between an Instant model and a Thinking model, you set how much reasoning ChatGPT applies to a given response: minimal for a quick lookup, higher for research, long-form writing or code.

This applies to ordinary ChatGPT conversations. The models behind ChatGPT Work and Codex are not part of this update.

Who this reaches, and when

OpenAI did not name any country restrictions in its announcement — the gate is your account tier, not your location — and it says a billion people now use ChatGPT each week. The updated Sol model was described as available immediately for paying subscribers, while the free-tier pieces roll out in two stages.

If you want to know where you stand, open the model picker at the top of a chat: Free and Go accounts will show Luna once the switch reaches them. There is no toggle to pull either change forward, and updating the mobile app will not force it, since both are enabled from OpenAI's side.

Why it matters

Message caps have been the sharpest edge of the free tier for years, and the usual reason people either paid up or drifted to a rival assistant. Removing them on text — while holding the line on images, uploads and voice — pushes the paid tiers toward being about tools and media rather than about how much you are allowed to type.

It is a busy stretch for OpenAI's consumer lineup. The company recently confirmed that ChatGPT Atlas would stop working and that bookmarks would not transfer, and the assistant landscape on phones is shifting too, with Google Assistant winding down in September.