Halfway through a late-night research session, ChatGPT quietly swaps you onto a slower model — like a bartender switching to well liquor when you weren’t looking. That particular frustration is officially over. OpenAI has announced the removal of text chat caps for free users, replacing rolling quotas with unlimited text conversations and a new default model for Free and Go accounts. A reasoning control for harder questions is also part of the rollout. As ChatGPT approaches one billion weekly users — a figure OpenAI itself has referenced, though earlier third-party reporting cited 900 million as of February 2026 — this is less a minor feature update and more a platform committing to text chat as standard infrastructure.
What “Unlimited” Actually Means
The fine print here matters as much as the headline, so read it before celebrating.
Think of this like Spotify’s free tier: the core product opens up, but the premium features stay behind the velvet rope. Text chats are uncapped. The following remain subject to separate limits:
- File uploads
- Image analysis
- Voice tools
- DALL-E image generation
If your typical session involves writing help, quick research, or Q&A, the friction is gone. Drop a PDF or request a generated image, and you’re still counting.
OpenAI is introducing a new default model for Free and Go users — a lightweight variant in its latest model family, built for general conversation and everyday tasks. According to OpenAI’s internal testing as reported by 9to5Mac, factual errors are significantly reduced compared to the previous GPT-5.5-Instant baseline, suggesting a meaningful quality improvement alongside the access expansion. The free version, in other words, didn’t just get more generous. It reportedly got smarter.
Rollout is staggered: the new default model goes live first, with unlimited text conversations and a new Think button following for Free and Go users shortly after.
What Paying Users Get (And Why It Still Matters)
Plus and Pro subscribers receive a distinct model upgrade and a new tool for controlling reasoning depth.
Plus and Pro users gain access to an upgraded model variant — separate from the versions powering Codex and Work products — that OpenAI describes as delivering more compact, robust answers with improved reliability. Per 9to5Mac’s reporting of OpenAI’s announcement, internal testing showed substantially fewer factual errors compared to GPT-5.5-Instant. The upgraded model and a new thinking slider are available to subscribers as part of the same rollout.
That reasoning control deserves a plain-language explanation. It works like the difference between firing off a quick text and actually drafting a considered reply — like deciding whether to leave someone on read or take a beat and compose something worth sending. Free and Go users get a Think button that prompts the model to slow down on complex questions. Plus and Pro users get a slider to dial reasoning depth based on query complexity, giving them finer control over the speed-versus-thoroughness trade-off.
OpenAI is clearly betting that text chat becomes infrastructure — unmetered, ambient, and always available. With unlimited text as the new baseline, the revenue logic shifts toward voice, multimodal tools, and deeper research capabilities, all of which remain gated by subscription tier. For the hundreds of millions of people using ChatGPT regularly, though, the free experience just received a genuine, no-asterisks upgrade.
