Artificial Intelligence

1 Billion Weekly Users, Free ChatGPT Goes Unlimited

OpenAI says ChatGPT has crossed 1 billion weekly users and is making GPT-5.6 Luna the free default with unlimited text chat, as Bloomberg reports a 2027 hardware device with Jony Ive's LoveFrom.

Artificial IntelligenceBy Servola Tech Desk2026-08-074 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

1 Billion Weekly Users, Free ChatGPT Goes Unlimited

Key takeaways

ChatGPT has crossed 1 billion weekly active users, and OpenAI's response was to strip text-chat limits from the free tier rather than push holdouts toward a paid plan: unlimited text chat rolls out to Free and Go users next week.
GPT-5.6 Luna becomes the default model for ChatGPT Free and Go users this week, with unlimited text conversations and a new Think button for harder questions following in the week after; uploads, images, voice and other tools stay capped.
OpenAI's own internal evaluations found factual errors were 62 percent less common with GPT-5.6 Luna and 68 percent less common with GPT-5.6 Sol than with GPT-5.5-Instant, the accuracy claim behind this week's model swap for every paying and non-paying tier.
Bloomberg reported OpenAI is developing a donut-shaped speaker with Jony Ive's studio LoveFrom, priced at 300 to 400 dollars for a 2027 release, arriving while OpenAI is still fighting Apple's trade-secrets lawsuit over the hardware team it hired away.

A Billion Weekly Users, Then the Free Tier Opened Up

OpenAI said in its own post that ChatGPT has crossed 1 billion weekly active users, covering everything from quick questions to complicated decisions, and it picked that milestone as the moment to make the biggest free-tier change since launch. Starting this week, GPT-5.6 Luna becomes the default model for everyone on the Free and Go tiers, replacing GPT-5.5 as the model people get when they open a new chat without picking one themselves.

What changes, and when: GPT-5.6 Sol's upgrade for Plus and Pro subscribers is available today. GPT-5.6 Luna becoming the free default lands this week. Unlimited text-based conversations and a new Think button, which lets a user request higher reasoning power for a harder question, are both set to roll out to Free and Go users next week, according to OpenAI's own announcement.

The unlimited part is scoped narrowly. Text chat loses its message ceiling, but OpenAI says separate limits stay in place for file uploads, images, and other tools, so a free account trades one kind of cap for a different, still-present set of them.

The Accuracy Number Behind the Model Swap

The evaluation OpenAI is pointing to: the company's own internal testing found factual errors were 62 percent less common with GPT-5.6 Luna and 68 percent less common with GPT-5.6 Sol than with GPT-5.5-Instant, the model both are replacing. That is the number OpenAI is using to justify handing its newest model to every tier at once rather than holding it back as a paid-only upgrade.

Plus and Pro subscribers get a thinking slider that adjusts how much reasoning the model applies to a given question. Free and Go users get a simpler on-off Think button instead, which is a smaller control surface but the same underlying reasoning capability that used to sit entirely behind a subscription.

The Real Story Is Distribution, Not Monetization

Read the sequencing: a company does not usually remove the one limiter that nudges casual users toward a paid plan unless it has decided scale matters more than near-term revenue from that group. OpenAI just did exactly that, in the same announcement where it disclosed passing 1 billion weekly users, and handed the free tier its newest, most accurate model rather than an older one. That is a land-grab-now, monetize-later sequence: lock in distribution and habit while the model is good enough that switching away from ChatGPT costs the user something.

The hardware wrinkle: Bloomberg reported this week that OpenAI is developing its first hardware device with Jony Ive's design studio LoveFrom, a donut-shaped, screenless speaker roughly the size of a hockey puck, expected to cost 300 to 400 dollars and to reach the market in 2027. Nothing about it is confirmed by OpenAI itself, no specs, no launch date beyond the general 2027 window, and it should be read as a report, not a product announcement.

The timing still matters. OpenAI is simultaneously defending a trade-secrets lawsuit in which Apple alleges it hired away people with knowledge of Apple's own hardware manufacturing processes, a dispute Servola has covered separately as it developed. A software land-grab that costs OpenAI nothing to ship is a rational move to make while the far harder, and currently contested, hardware bet is still years from shelves.

What This Means for EU and UK Readers

No separate gate, for now: neither OpenAI's announcement nor the coverage of it mentions any region-specific rollout, waiting list, or approval step, so Free and Go users in the EU and UK should expect GPT-5.6 Luna, unlimited text chat, and the Think button on the same weekly schedule as users everywhere else.

Where EU rules would actually bite is hardware, not this update: this week's change is software running on servers OpenAI already operates under existing data protection obligations. A physical donut-shaped speaker with an always-on camera and microphone array sold into the EU or UK is a different category of product entirely, one that would need CE marking and compliance with the EU's product-safety framework, plus the kind of privacy-by-design and data protection impact assessment scrutiny that a chat app in a browser tab never faces. None of that is triggered yet, because the device itself is still unconfirmed, but it is the gap worth watching once OpenAI, or Bloomberg's sourcing, produces an actual specification sheet.

Frequently asked questions

Will the unlimited free-tier text chat apply in the EU and UK immediately?

Nothing in OpenAI's announcement or the reporting around it mentions a region-specific rollout, so EU and UK Free and Go users should see GPT-5.6 Luna and, the week after, unlimited text chat and the Think button on the same schedule as everywhere else. If that changes, it has not been announced yet.

What still has a limit once ChatGPT's free text chat is unlimited?

File uploads, image generation, voice features, and other tools keep their existing per-tier limits. Only back-and-forth text conversation loses its message ceiling, and even that change had not shipped as of this announcement, it was described as arriving the following week.

Has OpenAI actually confirmed the donut-shaped speaker?

No. Every detail, the shape, the 300-to-400-dollar price band, the 2027 timing, LoveFrom's involvement, comes from Bloomberg's reporting rather than an OpenAI announcement. OpenAI has not published specifications or a launch date, and the device should be treated as a credible report in development, not a confirmed product.

The number worth remembering from this week is 1 billion, not 5.6. OpenAI did not respond to hitting that milestone by asking casual users to pay more, it responded by removing the one thing that used to push them toward paying at all. That is a company still optimizing for reach, running its most capable free model out to everyone at once, while a much harder and still-unconfirmed hardware bet plays out years ahead and, right now, in a courtroom.

OpenAIChatGPTGPT-5.6free tierunlimited chathardwareApple lawsuitJony Ive

Dropping the Name Did Not End the Lawsuit

OpenAI stopped using the io brand months ago and still settled with iyO this week. The concession arrived after the complaint had been amended.

5 min read

Infrastructure

OpenAI Paid You Back in the Thing That Broke

Four incidents on 27 July alone, on OpenAI's own status page. Compensation for the weekend outage was a usage reset, which is credit on the service that failed.

4 min read

Decisions

The Codename Worked. The Trading Controls Did Not

An indictment unsealed on 24 July says two Volkswagen engineers traded on a joint venture known internally as Project Climb. The control that failed is a common one.