OpenAI: Unlimited Free ChatGPT Text Chats + Think Button

OpenAI is making its biggest free-tier move of 2026: GPT-5.6 Luna becomes the default model for Free and Go users this week, replacing GPT-5.5, and unlimited text chats roll out next week alongside a new Think button that unlocks higher-reasoning mode.

The economics behind the move are stark. OpenAI’s internal evals put Luna’s factual-error rate 62% below GPT-5.5-Instant and Sol’s 68% below — the free tier is inheriting a markedly more accurate model at what the company is now treating as close-to-zero marginal cost.

Key facts

• New default: GPT-5.6 Luna replaces GPT-5.5 for Free and Go tiers; Sol-level reasoning stays gated behind the Think button and paid tiers.
• Unlimited text: the cap on text chats disappears for free users — the first time OpenAI has removed the free-text ceiling since usage limits were introduced.
• Think button: a manual “reason more” control that routes the request to deeper reasoning (Sol-class) rather than the default fast path.
• What’s still limited: file uploads, image generation, voice and image inputs keep their quota — text is the untapped layer, multimodal stays metered.
• Error-rate gap: internal evals claim Luna’s factual-error rate is 62% lower than GPT-5.5-Instant (Sol: 68% lower), the justification for letting the free tier run unrestricted.

Why it matters

• The free tier becomes the funnel, not the demo. Unlimited text + a smarter default converts ChatGPT from a taste-of-AI product into a daily utility for billions of users — and every free session is training data, distribution, and habit formation.
• Competition moves to the reasoning edge. With text unlimited, differentiation shifts to the Think button and solver-tier quality — the same frontier OpenAI is monetizing at the top end.
• It’s the end of the hard paywall era. Rivals (Google’s Gemini, Anthropic’s Claude) have been converging on the same shape: unlimited commodity text, premium reasoning behind a button — the limit is no longer “how many questions,” it’s “how much thinking.”

What to watch

• Latency and reliability at scale once the free text load goes flat-out — ChatGPT’s infrastructure has been OpenAI’s quietest differentiator.
• Adoption curves of Think-mode — if free users habitually tap Think, the gap between headline-free and actually-reasoning collapses.
• Mirror moves from Google and Anthropic — expect equivalents within weeks, not months.

Official source

• TechCrunch: OpenAI brings unlimited ChatGPT text chats to free users
• Context: AI Weekly — top AI stories, August 10, 2026

Updated August 10, 2026.