ChatGPT's 6 August 2026 update: what changed for paid and free users

On 6 August 2026, OpenAI pushed a notable update to ChatGPT that affects every tier of the product, though in quite different ways. If you pay for Plus or Pro, the model you rely on just got meaningfully more accurate and gained a new control for adjusting how hard it thinks. If you use the free tier, your default model is changing, and unlimited text conversations are coming your way. Here is what actually changed and what it means in practice.

What changed for Plus and Pro subscribers

The centrepiece for paid users is an improved version of GPT-5.6 Sol, the flagship model in OpenAI’s GPT-5.6 family. OpenAI says responses containing at least one factual error fell by 68% compared with GPT-5.5 Instant, based on internal evaluations covering financial, medical, and legal prompts. That is a significant claimed improvement, though it is worth noting these figures come from OpenAI’s own benchmarks rather than independent testing, and the methodology details, sample size and error definitions are not fully disclosed. Treat them as a directional signal rather than a verified measurement.

Beyond accuracy, the updated Sol is designed to be less verbose. Simple questions now get direct answers with enough context to be useful, rather than walls of bullet points and caveats. Complex requests still receive fuller responses, but the model is meant to lead with its actual recommendation rather than burying it at the end.

OpenAI has also adjusted the model’s behaviour around agreement. Rather than defaulting to validating whatever a user says, Sol will now offer a correction when simply agreeing would not actually be helpful. Small change, but a welcome one if you have ever caught ChatGPT enthusiastically confirming something incorrect.

The reasoning slider

The most visible change for Plus and Pro users is a new slider that controls how much reasoning effort ChatGPT applies. This replaces the previous system of separate Instant and Thinking modes with a single, continuous control. At the lower end, responses come back almost immediately and suit everyday questions. At the higher end, the model can spend several minutes working through complex problems, which is more useful for planning, research, or tricky coding tasks.

The slider is available across web, mobile, and desktop. It was already present in ChatGPT Work; this update brings it to the main Chat experience as well.

It is also worth knowing what this update does not touch: the version of GPT-5.6 Sol powering ChatGPT Work and Codex is unchanged. This release is specifically about the everyday chat experience.

What changed for Free and Go users

From 6 August 2026, OpenAI is switching the default model for free and entry-level Go accounts from GPT-5.5 Instant to GPT-5.6 Luna, the most cost-efficient member of the GPT-5.6 family. Luna is a smaller model than Sol, but it is not without substance: OpenAI reports a 62% reduction in factual errors for Luna compared with GPT-5.5 Instant, using the same internal evaluation methodology.

Alongside the model switch, OpenAI is rolling out unlimited text conversations for free users. The rollout is expected to complete for all eligible accounts within the week following 6 August 2026. File uploads, image generation, voice chat, and other tools remain subject to limits.

There is also a new Think button for free and Go users. Tapping it gives Luna extra time to reason through harder questions before responding. It does not switch you to a more powerful model; it simply lets the smaller Luna work longer on a given problem.

The trade-off worth knowing about

That last point matters. The Think button extends Luna’s reasoning time, but free users no longer have access to Sol’s reasoning capabilities at all. Critics have noted this is a meaningful step back from what was previously available. If you relied on deeper reasoning from ChatGPT without paying, the Think button partially offsets that loss, but it is not an equivalent replacement. It is a fair trade to understand before assuming unlimited chats means unlimited capability.

Why OpenAI is making these changes

A few things are happening at once. ChatGPT has reached one billion weekly users, and OpenAI has said that milestone is prompting it to rethink how much capability it makes available at each tier. On 30 July 2026, the company reduced the price of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%, which makes it more economically viable to offer Luna to free users at scale.

The reasoning slider is part of a longer-term design direction: rather than presenting thinking and non-thinking as separate, noticeably different models, OpenAI wants them to feel like different operating modes of one coherent system. That is a more intuitive experience for most people, and it removes some of the confusion about which model to pick.

What this means for you

If you are a Plus or Pro subscriber, the update is straightforward good news. More accurate answers, less padding in responses, and a simpler way to dial up reasoning when you need it. The 68% accuracy improvement claim is OpenAI’s own figure, so apply appropriate scepticism, but the directional improvement appears genuine based on early hands-on coverage.

If you are on the free tier, you are getting a meaningfully better default model and the convenience of unlimited text conversations, which removes one of the more frustrating friction points of the free experience. The Think button gives you a way to get more thoughtful responses on hard questions. The caveat is that the ceiling is lower than before: Sol’s reasoning is now a paid feature, full stop.

For developers and enterprise buyers, the update reinforces a pattern worth watching. Providers are increasingly using model tier separation to differentiate paid access, not just in speed or limits, but in the quality of reasoning available. As that gap widens, the decision about which plan is appropriate becomes a meaningful one for anyone using ChatGPT for serious work.