GPT-5.6 Sol’s new ChatGPT tuning is a product split, not an API upgrade

OpenAI has updated GPT-5.6 Sol inside ChatGPT for Plus and Pro users. The company says the change is intended to produce more focused answers and improve factual reliability, while a new slider lets a user choose how much thought ChatGPT applies to a response. OpenAI also says the same ChatGPT model now covers both Instant responses and deeper reasoning, positioning the change as a more continuous interaction model rather than a separate reasoning product.

The timing matters because this is a follow-on product adjustment, not the July 30 GPT-5.6 API and pricing update. In that earlier announcement, OpenAI described faster GPT-5.6 Sol processing through Fast mode and lower prices for the Terra and Luna models. The August 6 announcement draws a firm boundary around the latest change: the ChatGPT-optimized GPT-5.6 Sol is only available in the Chat experience, and the GPT-5.6 Sol used by Work and Codex is not changing as part of it.

OpenAI’s stated comparison point is GPT-5.5 Instant. In an internal evaluation of financial, medical, and legal prompts that required factual detail, it reports that answers with at least one factual error were 68% less common for GPT-5.6 Sol than for GPT-5.5 Instant. That is a useful directional claim about the intended improvement, but it is vendor-reported evidence rather than an independent evaluation, and it does not establish the error rate that a particular production workflow will see.

For technical users, the immediate practical change is in interactive ChatGPT work: the slider can be used to trade speed for more deliberation on planning, research, writing, coding, or decisions. It should not be treated as evidence that API calls, Codex workloads, or ChatGPT Work flows changed, because OpenAI explicitly excludes those surfaces from this update. Free users are instead being moved toward GPT-5.6 Luna, with unlimited text chats and a Think button rolling out on the schedule OpenAI described; uploads, images, and other tools remain subject to limits.