OpenAI is tearing down one of ChatGPT’s most visible free-tier barriers just as the market is making frontier AI harder to sell at a premium. The giveaway is broad, but the most powerful controls remain behind the paywall.

The rollout begins this week with GPT-5.6 Luna becoming ChatGPT’s default model for all users. OpenAI says Luna, built for speed, produced responses with at least one factual error 62% less often than the prior GPT-5.5 Instant default.¹

Next, free users and subscribers to the low-cost Go plan will receive a “Think” button for tougher questions. Then, starting next week, those tiers will get unlimited text chats with Luna—ending the rate limits that had constrained frequent use. The change applies only to text: uploads, images and other multimodal features will still be capped.²

For paying Plus and Pro customers, OpenAI is taking a different tack: not more basic access, but greater control. GPT-5.6 Sol is being updated for more reliable, direct answers, while a new slider will let subscribers choose “how much thought ChatGPT puts into an answer.”² Sam Altman framed the package more bluntly: “5.6 Sol much better in chat now and unlimited text chat for free users!”Sam Altman (sama)³

That split captures the company’s commercial calculation. OpenAI presents the move as wider access to capable AI; Greg Brockman said the company was bringing “unlimited text chats with Luna, a quite capable model, to free users.”Greg Brockman (gdb)⁴ But the timing also points to a more defensive reality. Cheaper Chinese open-weight models have increased pressure on OpenAI and Anthropic, and OpenAI recently cut prices for Luna and Terra only weeks after their launch—far faster than the industry’s usual pricing cycle.¹

Free users get a bigger doorway. Paying users get the knobs. OpenAI’s challenge is ensuring that distinction still feels worth paying for.