ChatGPT Removes Text Chat Limits for Free Users, Adds New GPT-5.6 Models

OpenAI announced on August 6, 2026 that it is removing limits on text-based chats for all ChatGPT users, including those on the free tier. The change comes as ChatGPT crossed 1 billion weekly users. Free and Go users will move to a new GPT-5.6 Luna default model, while Plus and Pro subscribers get an upgraded GPT-5.6 Sol and a thinking slider. According to OpenAI's internal evaluation, both new models cut factual errors significantly compared to the previous GPT-5.5-Instant baseline.

What happened

Detail	Fact
Announcement date	August 6, 2026
ChatGPT weekly users	1 billion (recently crossed)
New default model for Free and Go	GPT-5.6 Luna (replaces GPT-5.5)
New model for Plus and Pro	GPT-5.6 Sol (upgraded version)
Factual error reduction vs GPT-5.5-Instant: Luna	62% fewer errors
Factual error reduction vs GPT-5.5-Instant: Sol	68% fewer errors
Rollout for Free and Go users	This week, unlimited text and Think button next week
Plus and Pro Sol availability	Available today (August 6)

OpenAI is dropping the text conversation cap for free and Go ChatGPT users. The new default model powering those conversations is GPT-5.6 Luna. Both tiers also get a “Think” button, which lets users request higher reasoning effort for complex questions.

Limits on files, images, voice messages, and image generation remain separate. Removing those is not part of this update.

What changes for paid users

Plus and Pro subscribers are getting an upgraded version of GPT-5.6 Sol, described by OpenAI as better suited for quick tasks: questions, web research, writing, planning, and decisions. The company says the model produces more compact and direct answers.

Worth noting: this upgraded Sol is separate from the GPT-5.6 Sol used inside Codex and Work, which is unchanged.

Paid users also get a thinking slider. Rather than a binary on/off, the slider lets them dial in how much reasoning the model applies based on how complex a query is and how many steps it involves.

Why it matters

Removing the text cap is a meaningful competitive move. Free users now get a daily-driver experience without hitting a wall mid-conversation, which lowers the friction for anyone who has been sitting on the fence about signing up or upgrading.

The accuracy numbers are the more interesting headline for business users. A 62-68% reduction in factual errors versus GPT-5.5-Instant is a large claim, sourced from an internal evaluation, so it should be treated as directional rather than gospel. Still, if it holds up in practice, it narrows the gap between the free tier and paid outputs for everyday tasks like research drafts, email writing, and data lookups.

The thinking slider for Plus and Pro is a practical quality-of-life improvement. Right now users either get standard output or flip to full reasoning mode. A granular slider means less wasted compute on simple queries and more depth reserved for the ones that actually need it. This kind of fine-grained control over AI reasoning is also relevant to anyone building on top of the API, as it suggests OpenAI is moving toward exposable reasoning parameters. If you’re exploring AI integration for your business workflows, that flexibility could matter.

Our take

Unlimited free text chats is a good headline, but the real story is model quality. OpenAI is shipping multiple named model variants (Luna for free tier, Sol for paid, a separate Sol for Codex and Work) that are tuned for different use cases. That segmentation is getting complex fast, and it’s worth paying attention to which model you’re actually hitting when you build automations or prompts that rely on consistent behavior.

The factual error numbers come from OpenAI’s own internal tests, not a third-party benchmark. We’d want to see that validated on real-world tasks before shifting any production workflows. That said, the direction is clear: OpenAI is raising the floor on free-tier quality while giving paid users more control. For most small business operators who use ChatGPT for content, research, or planning, the free tier is about to get noticeably better.

If you’re already using ChatGPT inside a broader workflow, check out our coverage of how Microsoft is routing GPT-5.6 as the default in Copilot, since model changes upstream can affect tools you may already depend on.

What to do about it

1. Test GPT-5.6 Luna this week on tasks where you previously hit the free-tier message cap, and note whether output quality on factual queries improves.
2. If you’re a Plus or Pro user, explore the thinking slider on a few representative prompts before setting a default position. Match reasoning depth to task complexity to avoid slower, costlier outputs on simple requests.
3. If you have prompts or automations built around GPT-5.5 behavior, run a comparison against GPT-5.6 before updating anything in production.
4. Hold off on treating the 62-68% error-reduction claim as validated until independent benchmarks confirm it.

The free tier just became a lot more usable. Test it now, but keep your production prompts pinned to a specific model version until you’ve confirmed the behavior change works in your favor.

Source: TechCrunch · AI