OpenAI Refreshes GPT-5.6 Sol and Opens Unlimited Text Chat to Free Users

OpenAI on August 6 refreshed the model that greets more than a billion weekly ChatGPT users, collapsing its "instant" and "thinking" experiences into a single upgraded GPT-5.6 Sol for paying subscribers and, in the more consequential move, handing free users a smaller model called GPT-5.6 Luna paired with unlimited text chats. The update reframes what a free AI account is worth, and it lands as rivals still gate their best chat models behind rate limits and paywalls.

For Plus and Pro subscribers, the new Sol now powers both quick replies and deeper reasoning, ending the older split in which fast answers and hard problems felt like two different assistants with different tones. "For Plus and Pro users, the same model now powers both Instant responses and deeper reasoning, creating one consistent experience," OpenAI wrote. A new reasoning-depth slider, available on web, mobile, and desktop, lets paid users dial effort up for planning, research, writing, or coding, or keep it quick for everyday questions.

What changed, and for whom

The headline claim is accuracy. OpenAI said that in an internal evaluation of financial, medical, and legal prompts requiring factual detail, responses containing at least one factual error were about 68% less common with GPT-5.6 Sol and about 62% less common with GPT-5.6 Luna than with the older GPT-5.5 Instant. On its own account, OpenAI framed it bluntly: "In our high-stakes factuality evaluation covering finance, medicine and law, the new GPT-5.6 Sol produced 68% fewer responses with factual errors than GPT-5.5 Instant." The company also tuned Sol for tighter, more focused answers that adapt detail to the question and offer a correction "when simply agreeing wouldn't be useful."

Those figures deserve a caveat. OpenAI has not published the prompt set, the sample size, or the grading procedure behind the 62% and 68% numbers, and no third party has reproduced them. They are the company's own internal measurements, not independent benchmarks.

For Free and Go users, GPT-5.6 Luna becomes the default model this week, with unlimited text chats and a new Think button arriving the following week, subject to abuse guardrails. "For questions that require deeper reasoning, Free users can tap the new Think button to give GPT-5.6 Luna more time to work through the answer," OpenAI said. Crucially, only text chat is unlimited. Caps remain on file uploads, image generation, voice, and other tools. Some critics, including The Decoder, framed the change as steering free users toward OpenAI's weakest model even as the marketing emphasizes abundance.

The version of Sol shipping here is scoped to the consumer chat surface. OpenAI noted the Sol that powers its Work and Codex products is not changing as part of this release.

The economics: marginal cost toward zero

Strip away the model names and the story is about unit economics. OpenAI could not credibly offer unlimited text chat to free users a year ago; inference was too expensive. That it can now, using a smaller Luna model that still beats last generation's flagship instant model on the company's own factuality test, is evidence that the marginal cost of serving a competent text answer is sliding toward zero. The July 30 post on "advancing the price-performance frontier with GPT-5.6" set up exactly this: cheaper intelligence per token makes give-it-away tiers viable.

When the cost of an incremental chat approaches nothing, the strategic value shifts from metering usage to owning the default. Unlimited free text is a distribution play: keep a billion weekly users inside ChatGPT, harvest engagement and data, and upsell the higher-margin extras, file handling, image generation, voice, and the reasoning slider, that remain capped.

That puts direct pressure on rivals' paid tiers. Google's Gemini, Anthropic's Claude, and a field of challengers have leaned on message limits and model gating to protect subscription revenue. If OpenAI has normalized unlimited free text chat, "how many messages do I get" becomes a weaker upsell, and competitors must either match the giveaway, absorbing the cost, or justify why their paid plans are worth it on capability and workflow rather than access alone. OpenAI cast the move in mission terms: "This is a concrete step toward more abundant intelligence... letting free users keep text chats going without a rate limit."

The release also arrives with expanded protections for users believed to be under 18, including training to avoid romantic roleplay and age-appropriate boundaries on sensitive topics, detailed in an accompanying system card.

What to watch next

Three things. First, whether "unlimited" survives contact with real load, or whether quiet throttling appears once next week's rollout scales. Second, whether independent researchers can reproduce the 62% and 68% error-reduction claims, which remain unverified outside OpenAI. Third, the competitive response: watch Google and Anthropic for matching free-tier moves in the coming weeks, and watch whether OpenAI's paid conversion holds now that the free product does more than ever.