OpenAI removes text-chat limits for free ChatGPT users, debuts GPT-5.6 variants with reasoning controls

Unlimited text chats now live for free users

OpenAI has lifted message-count restrictions on ChatGPT’s text-based interface for free and Go subscribers, according to TechCrunch reporting on August 6. The update, rolling out this week, is powered by GPT-5.6 Luna, a new base model that replaces GPT-5.5 as the default for free-tier users. The move comes as ChatGPT recently surpassed 1 billion weekly active users, a milestone OpenAI announced alongside the feature expansion.

Non-text modalities retain rate limits. According to TechCrunch, users continue to face separate ceilings for file uploads, image inputs, voice interactions, and image generation. The asymmetry reflects a deliberate product-tier design, though OpenAI has not publicly disclosed the reasoning behind preserving these constraints.

GPT-5.6 variants and the new reasoning slider

OpenAI is deploying two versions of its latest model family. Free and Go users receive GPT-5.6 Luna as their default, while Plus and Pro subscribers gain access to GPT-5.6 Sol, optimized for rapid-response tasks such as web research, writing, and decision-making. According to TechCrunch, the Sol variant “will give more compact and robust answers” compared to Luna.

Both tiers also unlock a new “Think” button for complex queries. Plus and Pro users receive an additional feature: a reasoning slider that lets them adjust the depth of computation per question. The slider’s implementation suggests users can now trade inference latency against answer quality on a per-query basis, though OpenAI has not detailed the slider’s behavior or computational footprint.

The Sol model available to Plus and Pro users is distinct from the version used in Codex and Work products, according to TechCrunch; that variant remains unchanged.

Internal benchmarks show factual-accuracy gains

Compared to GPT-5.5-Instant, OpenAI’s internal evaluation reports 62% and 68% reductions in factual errors for GPT-5.6 Luna and Sol, respectively, according to TechCrunch. These gains are measured in absolute error reduction, not relative to other contemporary models. No independent third-party benchmarking of the new variants has been published as of the announcement date.

GPT-5.6 Sol for Plus and Pro subscribers launched today; Luna and the Think button arrive to free users later this week.

Why This Matters

Teams evaluating ChatGPT for production workflows can now baseline against unlimited text queries on the free tier, removing a previous friction point for adoption trials. For power-user segments (researchers, writers, developers), the reasoning slider introduces fine-grained control over compute allocation, enabling query-specific trade-offs between response latency and depth without tier switching. Organizations standardizing on ChatGPT will need to re-evaluate their subscription calculus, since the gap between free and paid text-generation capability has narrowed, though document and image handling remain tiered. The 62–68% factual-error reduction claims are OpenAI-reported; independent reproduction on public benchmarks will inform whether the gap holds across domains outside OpenAI’s test set.