ChatGPT brings unlimited text chats to free users
TechCrunch Ivan Mehta ● Covered by 6 sources
OpenAI is killing text chat limits for free ChatGPT users, rolling out a new GPT-5.6 Luna model next week. It also claims way fewer factual errors than before.
OpenAI just made ChatGPT's free tier a lot more generous. Starting next week, anyone using the free version of ChatGPT will get unlimited text-based conversations, no more hitting a wall mid-chat because you asked one too many questions. This lands right after ChatGPT crossed 1 billion weekly users, which is the kind of milestone that tends to make a company loosen its grip on the free product rather than tighten it.
The change comes bundled with a new default model, GPT-5.6 Luna, which replaces GPT-5.5 for both Free and Go tier users. Alongside it, those users get a "Think" button, essentially a manual override that lets you tell the model to slow down and reason harder on tougher questions instead of firing off a quick response. Limits still apply to file uploads, images, voice, and image generation, so this is specifically a text-chat unlock, not an everything-goes free-for-all.
Paying users aren't left out. Plus and Pro subscribers get an upgraded GPT-5.6 Sol model today, tuned for fast, everyday tasks like research, planning, and decision-making, with OpenAI promising tighter, more efficient answers. Worth noting: this Sol update is distinct from the version powering Codex and Work, which stays untouched for now. Plus and Pro users also get a thinking slider, a more granular version of the Think button, letting them dial reasoning effort up or down depending on how gnarly the question is.
The more interesting number here might be the accuracy claim. OpenAI says its internal testing found GPT-5.6 Luna produces 62% fewer factual errors than GPT-5.5-Instant, and GPT-5.6 Sol comes in at 68% fewer. Those are OpenAI's own figures, collected internally, so they deserve the usual grain of salt reserved for a company grading its own homework. But if even roughly true, it suggests the free tier is getting a real quality bump, not just a limits removal dressed up as one.
The rollout is staggered. Sol's upgrade for paid users is live now, the Luna switch and Think button for Free and Go users arrive this week, and unlimited chats kick in next week.
My take
Unlimited free chats sound generous until you remember OpenAI needs to keep pace with Gemini and Meta AI throwing free compute around too, so this is competitive pressure dressed up as generosity. The self-reported error-rate drop is the kind of stat that means something only once outside researchers can poke at it, so treat it as a marketing headline until proven otherwise. Still, giving free users a genuine reasoning toggle instead of gating that behind a paywall is the right instinct, and other labs should be taking notes.
Read more about this at: TechCrunch