--- title: "OpenAI Drops All ChatGPT Text Limits for Free Users" date: 2026-08-06 author: "Barry Elad" featured_image: "https://sqmagazine.co.uk/wp-content/uploads/2026/08/openai-drops-chatgpt-text-limits-for-free-users.jpg" categories: - name: "Artificial Intelligence" url: "/artificial-intelligence.md" tags: - name: "News" url: "/tag/news.md" --- # OpenAI Drops All ChatGPT Text Limits for Free Users OpenAI removed daily message limits for free ChatGPT users on August 6, 2026, weeks after the assistant crossed 1 billion weekly users. The change covers text chats only. ## Quick Summary – TLDR: - OpenAI lifted daily text-chat caps for every free ChatGPT account, leaving limits on images, files and voice untouched. - GPT-5.6 Luna replaces GPT-5.5 Instant as the default model for Free and Go users this week. - OpenAI internal testing measured factual errors as 62% less common in Luna than in GPT-5.5 Instant. - Plus and Pro subscribers gain a thinking slider that sets how much reasoning each answer gets. - ChatGPT passed 1 billion weekly users before the change, sharpening questions about heavy daily use. ## What Happened? OpenAI announced the removal alongside a model swap that touches every tier. Free and Go users move to **GPT-5.6 Luna**, the lightweight member of the newest model family, which replaces **GPT-5.5 Instant** as the default. Plus and Pro users get an updated GPT-5.6 Sol tuned for quick everyday tasks such as web research, planning and writing. The rollout runs in stages. The revised Sol reached Plus and Pro accounts on announcement day, the Luna default lands across free tiers this week, and [unlimited text chats plus a new “Think” button](https://openai.com/index/improving-gpt-5-6-sol-in-chatgpt/) arrive next week. OpenAI released error-rate figures from its own evaluations to justify the swap. Against GPT-5.5 Instant, the company measured factual errors as **62%** less common in Luna and **68%** less common in Sol. No outside lab has reproduced those results, and OpenAI has not published the evaluation set behind them, so the gains sit inside the vendor’s own [hallucination measurement](https://sqmagazine.co.uk/llm-hallucination-statistics/) framing. ## Which Limits Actually Disappear? Text is the only modality losing its ceiling. Image generation, file uploads, voice tools and vision requests keep their existing quotas, and those are the features carrying the heaviest per-request compute cost. A free account can now run an unbounded conversation and still hit a wall on its first uploaded PDF. The pricing ladder behind that line is unchanged. Go costs **$8** a month, Plus **$20**, and Pro **$200**, and paid tiers now sell control over reasoning depth instead of raw message volume. Volume was the pitch when [GPT-5.5 shipped](https://sqmagazine.co.uk/gpt-5-5-faster-coding-research-ai/) as the speed-focused default. ## Paid Tiers Trade Caps for Reasoning Control Plus and Pro users receive a slider that dials thinking up from **GPT-5.6 high to extra high to pro**, matched to how many steps a query involves. Free and Go users get a coarser version of the same control in the “**Think**” button, which pushes one hard question into higher reasoning. OpenAI says the updated Sol returns more compact answers with less filler and formatting. OpenAI held one Sol variant out of the update. The GPT-5.6 Sol running in Codex and Work stays as it was, so developer and enterprise behavior does not shift while the consumer model does. That split matters for anyone [running agents on OpenAI infrastructure](https://sqmagazine.co.uk/openai-agents-sdk-sandbox-scalable-tools/) who benchmarked against the consumer endpoint. ## Safety Questions Follow the Unlimited Model Removing the cap removes the only structural brake on prolonged use. OpenAI states its models are built to “**avoid romantic roleplay, age-restricted challenges and presenting itself as a substitute for real-world relationships**” for users under 18. The company has not shown whether those guardrails were retested against sessions that now have no length ceiling. Litigation already covers this ground. **A mother’s lawsuit against OpenAI** alleges the company failed to enact safeguards and instead kept her daughter engaged with ChatGPT, echoing negligence claims against social media firms over deliberately addictive design. OpenAI’s own [pause on adult chatbot features](https://sqmagazine.co.uk/openai-pauses-adult-chatbot-safety-concerns/) showed the company moving carefully on engagement-heavy modes. Four questions stay open after the announcement: - **Whether the under-18 guardrails were re-evaluated for uncapped session lengths?** - **How many free accounts were hitting the daily cap before it was lifted?** - **Whether Luna’s error-rate gains hold outside OpenAI’s internal evaluation?** - **What happens to service quality if free demand spikes past current capacity?** Anyone who administers a teen or school account should confirm the age setting and parental controls before the cap lifts next week, since the change reaches existing accounts without a prompt. Doing so helps reduce exposure to long unsupervised sessions. ## SQ Magazine’s Takeaway The uncapped tier reads as a distribution decision made affordable by a cheaper default model. Luna is the lightweight option in the **GPT-5.6 family**, and swapping it in across the free base is what makes unbounded text serving viable at 1 billion weekly users. The limits that survive, on images, files and voice, mark exactly where OpenAI’s marginal cost per request is still high enough to meter. **What comes next is a two-week test of that math**. The Luna default lands this week and the cap removal follows next week, so the first signal to watch is whether response quality or availability slips under the new load. The second is regulatory: an assistant with no message ceiling and a documented under-18 policy invites scrutiny of whether the policy survives contact with sessions that never have to end.