ChatGPT Brings Unlimited Text Chats to Free Users

OpenAI announced on August 6, 2026 that it is lifting all text-chat message limits for ChatGPT's free and Go tiers — a notable concession for a platform that has long used those rate limits as the primary lever pushing users toward a paid subscription. The move arrives alongside a new default model, GPT-5.6 Luna, and a set of reasoning controls that signal a broader shift in how AI companies think about differentiating tiers.

What's Converging

The competitive pressure on AI companies to give away more has been building steadily throughout 2026. Google expanded Gemini's free context window earlier this year. Anthropic has moved Claude's free tier closer to parity with its paid offering in terms of daily availability. The pattern is consistent: as frontier model serving costs fall and the race for daily active users intensifies, the "hit your limit, upgrade now" friction becomes a liability rather than a feature. A platform that routinely turns users away trains them to seek alternatives.

OpenAI's own trajectory reinforces this. The company recently disclosed that ChatGPT has crossed 1 billion weekly users — a number that, if accurate, makes it one of the most widely used software products on the planet. Sustaining that scale requires keeping casual and free-tier users engaged, not bouncing them off a rate-limit wall. The decision to remove text-chat caps is less surprising in that light; it is more a maintenance move on retention than a sudden generosity pivot.

The third signal worth tracking is the commoditization of reasoning controls. Features like step-by-step thinking and extended inference time — once exclusive to premium, slower model variants — are now being surfaced as per-query toggles. OpenAI's new "Think" button and "thinking slider" are the clearest sign yet that reasoning depth is becoming a user-configurable dial rather than a fixed model property. That shift has real downstream implications for how AI products are designed.

The Specific Development

The centerpiece of OpenAI's August 6 announcement is straightforward: starting next week, users on the free and Go ($8/month) tiers can send unlimited text messages to ChatGPT. What is not changing is equally important — limits on image generation, file uploads, voice interactions, and image analysis remain in place. Multiple reports confirm this boundary explicitly. The Verge, TechCrunch, and CNET all note that the unlimited offer applies strictly to text-only conversations, which is the feature that cost the least to extend and affects the broadest share of everyday users.

The new default model for those tiers is GPT-5.6 Luna, which replaces GPT-5.5 and is described by CNET as the lightweight member of OpenAI's newest model family. OpenAI's internal evaluation claims 62% fewer factual errors for Luna compared to GPT-5.5-Instant — a significant margin, though internal evals should be read with the usual caveats about optimistic self-reporting. Free and Go users are also getting a "Think" button, a binary per-query toggle that routes harder questions through extended reasoning. That feature arrives next week, on the same timeline as the unlimited text-chat rollout.

For Plus and Pro subscribers, the update is different in character. They receive an upgraded version of GPT-5.6 Sol positioned as better suited for quick tasks — questions, research lookups, writing, planning, and decisions — with what OpenAI describes as more compact and robust answers. The accuracy claim here is 68% fewer factual errors versus GPT-5.5-Instant, slightly higher than Luna's figure. One detail that multiple sources take care to flag: this GPT-5.6 Sol is a chat-optimized variant, and it is explicitly separate from the GPT-5.6 Sol used inside Codex and Work, which is unchanged by this update. That bifurcation matters for developers who have been tracking Sol's behavior in agentic contexts — the model that powers chat and the one powering code-execution agents are now on different update tracks.

Our read is that the reasoning-slider feature for Plus and Pro is the most interesting UX bet in this package. Rather than forcing users to pick between a "fast" and a "smart" model at the start of a session, OpenAI is letting them tune per-query how much inference compute to spend. A planning question gets more thinking; a quick factual lookup gets less. That granularity shifts the cognitive overhead onto the user in a way that could either feel empowering or exhausting depending on the interface execution — something worth watching as it rolls out.

What's Likely Next

The immediate question is whether removing text limits meaningfully changes ChatGPT's competitive position, or whether the friction that mattered was always on richer modalities — images, voice, file analysis — which remain capped. If a significant share of churned free users left because of text rate limits specifically, OpenAI should see engagement numbers move in the next 30 days. If they were leaving because they wanted image generation or document analysis without hitting walls, this announcement does nothing for them, and a second round of limit relaxation on those features would follow.

The 90-day question is about paid tier differentiation. Unlimited text was one of the clearest reasons to upgrade from free to Plus. With that removed, OpenAI is betting that the thinking slider, Sol's quality improvement, higher caps on images and files, and priority access are sufficient to justify $20 a month. Competitors — Google, Anthropic, and the open-source ecosystem — will be watching whether Plus subscriber growth holds or softens in response. If it softens, expect OpenAI to either restore some limit (unlikely given the PR cost) or accelerate new paid-only features to rebuild the value gap.

Sources