ChatGPT Drops Chat Limits, Upgrades Its Core Model Now

Remember the little prompt that begged you to upgrade when your conversation hit a cap? Those moments are shrinking into memory. Today OpenAI announced that free users — and subscribers on the cheapest Go tier — will soon be able to run unlimited text chats without being silently downgraded to an older model.

Free and Go accounts will default to GPT‑5.6 Luna, replacing the long-serving GPT‑5.5 Instant. Luna, the smallest of the GPT‑5.6 family unveiled in July, is set to roll out as the default sometime this week. Unlimited text chatting follows next week, and that’s also when a new Think button will appear for free and Go customers, giving people control over how much “thought” the assistant expends on a reply.

Don’t get ahead of yourself — there are still constraints. File uploads and image generation will continue to have limits for nonpremium users. This change is focused squarely on the back-and-forth text experience, the place where most people live when they ask quick questions or brainstorm ideas.

On the paid side, OpenAI is refining GPT‑5.6 Sol, the flagship model reserved for Plus and Pro tiers. The update aims to make answers more direct, trim distracting formatting, and avoid dumping unnecessary detail into short Q&A. In other words: a straight answer when that’s what you need, and a fuller, structured reply when you’re planning, researching, or building.

The company also says Sol now makes fewer factual mistakes — especially around numbers, dates, sources, and the kind of rule-based logic that trips up language models. In internal tests using financial, medical, and legal prompts that require precision, answers with at least one factual error dropped by about 68% compared with GPT‑5.5 Instant. That’s a meaningful improvement for anyone relying on the model for work that can’t tolerate sloppy details.

Plus and Pro users will get a more granular Think control: a slider that ranges from “instant” to “pro,” letting you dial how much computation the model applies while keeping its tone consistent. Use it for everyday quick answers, or crank it up when you need planning, analysis, code review, or polished writing. One caveat: the variant of GPT‑5.6 Sol used in ChatGPT Work and Codex won’t change with this update; the new Sol is optimized for the Chat experience and focused on everyday interactions.

This feels like an attempt to rebalance access — more casual users get a better chat baseline, while paying customers receive a tidier, less error-prone flagship. Will lowering friction around text conversations change how people use AI day to day? Time will tell, but for now the chat box just got a little roomier.