Key Takeaways

  • Free and Go users lose all caps on text conversations and receive a new Think button for deeper reasoning.
  • The default model shifts to GPT‑5.6 Luna, which cuts factual errors by 62 percent versus the previous release.
  • Plus and Pro tiers get an upgraded GPT‑5.6 Sol and a thinking slider that lets them dial reasoning intensity per query.
  • Unlimited text chat rolls out next week; file, image, voice and generation limits stay unchanged.

OpenAI has thrown the doors wide open for its biggest audience. After crossing one billion weekly users, the company announced that every free and Go account will now enjoy unlimited text chats. The move eliminates the last major friction point that kept casual users from treating ChatGPT as a daily workhorse. No more counting messages, no more waiting for a reset — just type and keep typing.

The engine behind this openness is GPT‑5.6 Luna, a freshly minted model that becomes the default for the two lowest tiers. Internal testing shows Luna makes factual mistakes 62 percent less often than GPT‑5.5‑Instant. That is a concrete quality jump, not a marketing flourish. For users who have lived with occasional hallucinations, the improvement feels tangible.

A new Think button sits beside the prompt box. When a question demands more than surface‑level retrieval — multi‑step logic, code debugging, nuanced analysis — the button summons a higher‑reasoning mode. Free users get a single toggle; Plus and Pro users receive a continuously adjustable thinking slider. The slider lets paying customers decide exactly how much compute the model spends on a given answer, trading latency for depth on a per‑query basis.

Plus and Pro subscribers are not left out of the model upgrade. They inherit GPT‑5.6 Sol, a sibling of Luna tuned for speed and conciseness. The company claims Sol delivers more compact, robust answers for quick tasks such as web research, planning, writing and decision support. Internal benchmarks put Sol’s factual error rate 68 percent below the old instant model. The Sol variant used for Codex and Work remains untouched, preserving the specialized behavior developers rely on.

The rollout is staggered. Starting today, Plus and Pro users already see the upgraded Sol and the thinking slider. Free and Go users will receive Luna and the Think button this week. Unlimited text chat for those tiers arrives next week. File uploads, image analysis, voice interaction and image generation keep their existing caps — OpenAI is not handing out a blank cheque for every modality.

Strategically, the unlimited text tier is a land‑grab. By removing the message ceiling, OpenAI turns ChatGPT into a default writing companion for students, marketers, coders and anyone who needs a persistent drafting partner. The Think button adds a low‑friction lever for deeper work without forcing a subscription. Meanwhile, the thinking slider gives power users a granular control that competitors have only offered via opaque “max tokens” settings.

Critics will note that the quality gains, while measurable, still leave room for error. A 62‑percent reduction does not equal zero hallucinations. The Think button’s single‑press design for free users may feel coarse compared with the slider’s nuance. And the untouched limits on files and media keep the free experience from becoming a full‑blown multimedia studio.

Nevertheless, the signal is clear: OpenAI is betting that volume and accessibility will cement its lead before rivals can match the scale. Unlimited text chat for a billion‑plus user base is a data flywheel — more conversations, more feedback, faster model iteration. The company has turned a usage cap into a growth lever, and the market will feel the ripple.