OpenAI said on August 6 that GPT-5.6 Luna becomes the default chat model for ChatGPT’s Free and Go tiers, replacing GPT-5.5 Instant, with the swap rolling out over the announcement week. The same post schedules the end of text-chat rate limits for Free and Go accounts for the week of August 10.
What OpenAI Announced on August 6
In a post titled “Improving GPT-5.6 Sol in ChatGPT—and expanding access to GPT-5.6 Luna for free users,” OpenAI set out two changes for the free tiers: Luna replaces GPT-5.5 Instant as the model behind Free and Go conversations, rolling out across the announcement week, and text limits come off the week of August 10. OpenAI’s headline claim for Luna is accuracy: by the company’s own internal testing, Luna produces 62% fewer factual errors than GPT-5.5 Instant. That figure comes from OpenAI’s own testing, not an outside audit, so treat it as a vendor claim rather than a verified benchmark.
What Changes for Free ChatGPT Users?
Starting the week of August 10, Free and Go accounts no longer run into text-based rate limits inside ChatGPT chats, subject to anti-abuse limits OpenAI keeps in place. A new Think button lets a user ask Luna to spend more time on a harder question before it answers. File uploads, images, and other tools keep their existing caps; only text chat opens up.
Plus and Pro Get an Upgraded Sol
Paid tiers have theirs already: an upgraded GPT-5.6 Sol has been live for Plus ($20/month) and Pro ($200/month) accounts since August 6, with a reasoning-effort slider that, per The Next Web, runs from high to extra high to pro. OpenAI pitches the updated Sol, not Luna, as the reliability play: fewer mistakes when answers depend on dates, numbers, sources, rules, or assumptions, less unnecessary formatting, and a correction rather than agreement when agreeing wouldn’t be useful. Go is $8 a month. ChatGPT Work and Codex sit outside this update entirely. Sol has already come up once on this site: two of the 19 unsanctioned-activity cases in the AISI agent-evaluation incident traced back to a single run involving GPT-5.6 Sol, a separate story about testing infrastructure rather than about this rollout.
Why the Default Swap Matters for Visibility Tracking
Our read, not OpenAI’s: once the rollout lands, the model answering an average Free or Go prompt about a brand or product is a different one than the default in place before the August 6 announcement. A visibility snapshot taken before the announcement reflects the old default’s answers, and because the swap rolls out over a week, there is no single date on which every free account changed over. The rate-limit lift is scheduled for the week of August 10 and has not taken effect yet. That’s a reason to re-run brand-mention checks against the new default, not a claim that referral traffic or citation counts will move: none of the sources for this piece report a link between the model swap and either metric. Microsoft Clarity’s AI Visibility view, which separates branded from non-branded grounding queries behind AI citations, is one free dashboard for watching whether a site’s citation numbers drift over time.
The practical move is bookkeeping: date-stamp the last visibility check, treat anything gathered before August 6 as a reading of the old default, and re-run the same checks after the rollout week closes rather than assuming the older numbers still hold. Anyone tracking brand mentions through ChatGPT’s free tiers is watching a surface whose default model changed with this announcement, and the safest assumption is that the baseline moved with it.