OpenAI is updating GPT-5.6 Sol inside ChatGPT and widening GPT-5.6 Luna access for free users.
The August 6 product post says Plus and Pro users can use an updated GPT-5.6 Sol in ChatGPT starting now. The change adds a reasoning slider that lets users choose how much thought ChatGPT puts into a response, while OpenAI says the same Sol model now powers both instant and deeper-reasoning experiences for those plans.
The bigger distribution change is lower in the announcement. OpenAI says GPT-5.6 Luna will become the default model for Free and Go users this week. Starting next week, those users are slated to get unlimited text chats and a new Think button for harder questions, subject to abuse guardrails. File uploads, images, and other tools still have limits.
OpenAI also draws a boundary around the update. The company says this version of GPT-5.6 Sol is optimized for everyday ChatGPT conversations and is only available in the Chat experience. The Sol version that powers Work and Codex is not changing as part of this release.
The access layer is now part of the model story
This is a follow-up to OpenAI’s recent GPT-5.6 price-performance update, but the strategic center has shifted from API economics to default consumer access.
For paid ChatGPT users, the reasoning slider makes effort a visible product control. The useful promise is not simply “more reasoning.” It is letting a user spend more latency and tokens when the task is planning, research, writing, coding, or decision support, while keeping ordinary answers tighter.
For Free and Go users, Luna becoming the default matters because default models shape the median AI experience. If unlimited text chat lands as described, the constraint moves away from “how many messages do I have left” and toward where OpenAI still enforces limits: tools, files, images, and abuse protections.
Reliability claims need evaluation context
OpenAI says the new GPT-5.6 Sol is designed to be more focused and more reliable with facts. It also says that in an internal evaluation of financial, medical, and legal prompts requiring factual detail, responses with at least one factual error were about 62% less common with GPT-5.6 Luna and 68% less common with GPT-5.6 Sol than with GPT-5.5 Instant.
Those are useful directional claims, but they are not a substitute for workflow-specific testing. The model can improve on OpenAI’s internal factuality set and still fail on a company’s internal policy, a live legal rule, a narrow medical protocol, or a source that changed yesterday.
The practical read is to treat the update as a routing and access change first. Paid teams get a more controllable ChatGPT experience. Free users get a newer default model and more text-room to work. The hard work remains measuring whether those changes improve the actual tasks people repeat.
