OpenAI has removed the last visible barrier between free users and its current-generation models: the text chat cap. Announced on August 6, the change lets ChatGPT Free and Go accounts carry on unlimited text conversations, while the lightweight GPT-5.6 Luna becomes their default model, replacing GPT-5.5 Instant. Until now, free users hit rate limits of roughly 10 to 40 text messages per three-to-five-hour window on flagship models, after which they were demoted to a lighter model mid-conversation. Those limits, and the demotions, are gone for text.
The shift is more than a generosity gesture. It is a statement about economics and competition. OpenAI says a billion people use ChatGPT weekly, and the free tier is the funnel that keeps that base attached, and out of the reach of Claude, Gemini, and a growing field of free alternatives. Removing caps on the entry tier, and doing it with a current-generation model, is the consumer equivalent of the API price cuts OpenAI made in July, when it took 80 percent off Luna’s API pricing and 20 percent off Terra’s. Text-only inference has apparently become cheap enough that OpenAI would rather spend it as a retention tool than meter it as a revenue gate.
The details matter for what they reveal about where the cost actually sits. Luna, the smallest and quickest of the three GPT-5.6 tiers that reached general availability on July 9, becomes the default on the free and Go tiers, and OpenAI’s internal testing, run on financial, medical, and legal prompts, found that roughly 62 percent fewer of Luna’s responses contained a factual error compared with GPT-5.5 Instant. Free and Go users also receive a Think button, which lets Luna spend extra processing time on difficult questions. But the caps did not disappear; they migrated. Limits still apply to file uploads, images, voice mode, and image generation, which is to say the modalities that actually burn compute and money. The free tier is now unlimited exactly where it is cheapest for OpenAI to serve, and metered where it is not.
Paying customers get their own upgrades. Plus and Pro subscribers get a retuned GPT-5.6 Sol, tuned for everyday tasks like research, planning, writing, and decision-making, with OpenAI claiming about 68 percent fewer factual errors than Instant and more focused, less padded answers. A new slider lets them dial the model’s reasoning effort up or down per request. Notably, the refreshed Sol is available only in ChatGPT’s chat view; the Sol build behind Work and Codex is unchanged. The rollout is staggered: Sol and the slider arrived on August 6, the Luna default takes effect this week, and unlimited text plus the Think button follow next week, with abuse safeguards attached.
Two questions hang over the announcement. One is safety. OpenAI paired the change with under-18 protections described in a new system card, but the wider exposure is straightforward: an unlimited, always-on free chatbot is more time in front of vulnerable users, and lawsuits already filed against OpenAI claim that heavy chatbot use has harmed some of them. The other is what the cap’s relocation says about the industry’s cost curve. When a company with a billion weekly users decides that unlimited text is the way to hold them, it is a quiet admission that the marginal cost of text generation has fallen far enough to be given away, and that the next competitive battleground, where the money actually is, is everything that isn’t text.
Sources: Improving GPT-5.6 Sol in ChatGPT, and expanding access to GPT-5.6 Luna for free users (OpenAI, Aug 6, 2026); OpenAI is finally killing ChatGPT’s text chat limits for free users (PCWorld, Aug 7, 2026); ChatGPT brings unlimited text chats to free users (TechCrunch, Aug 6, 2026); OpenAI removes the daily limit on free ChatGPT chats and upgrades the default model to GPT-5.6 Luna (The Next Web, Aug 6, 2026); Free ChatGPT Users Get Unlimited Text Chats and GPT-5.6 Luna (MacRumors, Aug 6, 2026); OpenAI Gives Free ChatGPT Users Unlimited Text Chats on GPT-5.6 Luna (Unite.AI, Aug 6, 2026)