GPT-5.6 Sol's update is a ChatGPT behavior change, not a new Codex model

OpenAI updated GPT-5.6 Sol in the ChatGPT Chat experience for Plus and Pro users on August 6. The company says the revision is intended to give more focused answers, use less unnecessary formatting, adapt detail to the request, and better handle factual questions. It also adds a slider that lets eligible paid users choose how much reasoning effort ChatGPT applies to a response.

The timing matters because GPT-5.6 Sol had already begun rolling out to eligible paid ChatGPT plans on July 9. This is therefore a product-surface update to an existing model family rather than evidence of a separate model launch. OpenAI's current help documentation says Sol powers the Medium, High, and Extra High reasoning options on eligible plans, while Sol Pro powers the Pro option.

For technical users, the consequential boundary is between Chat and other OpenAI products. OpenAI explicitly says this Chat-optimized Sol version is available only in ChatGPT's Chat experience and that the Sol version powering Work and Codex is not changing. It is reasonable to treat the announcement as a change in interactive ChatGPT behavior and controls, rather than as evidence of altered coding-agent or API performance.

The comparison with the previous default is narrow. OpenAI reports that, in an internal evaluation of financial, medical, and legal prompts that required factual detail, answers containing at least one factual error were 68% less common with Sol than with GPT-5.5 Instant. That is a vendor-reported result about a specific factual-reliability evaluation, not an independent measurement of general coding, research, or agent performance.

Access remains constrained by plan and rollout. Free and Go users receive Luna rather than Sol, and their new Think control uses Luna with more reasoning time; OpenAI also says file uploads, images, and other tools remain subject to limits. The update may make paid ChatGPT responses feel more consistent across quick and deliberate interactions, but teams using Codex or the API should not assume those workflows changed without product-specific documentation or their own evaluation.