GPT-5.6 Sol’s latest change is confined to ChatGPT

OpenAI says it updated the ChatGPT-specific version of GPT-5.6 Sol on August 6 for Plus and Pro users. The stated changes are behavioral rather than a newly named model: more focused answers, tighter formatting, better adaptation to the question, and a reasoning slider that lets users select how much thought goes into a response. OpenAI also says the same Sol model now serves both Instant and deeper-reasoning interactions for those users, aiming to make a change in reasoning effort feel like more work by one assistant rather than a switch to a differently styled model.

The scope matters. In OpenAI's July 9 GPT-5.6 product post, Sol was presented as the flagship tier in a three-model family alongside Terra and Luna, with separate access routes across ChatGPT, Codex, and the API. The August 6 post narrows its own change explicitly to the Chat experience: it says the Sol version used by ChatGPT Work and Codex is not changing. That makes this a product-surface tuning pass, not evidence of a new API revision or a general capability change across every GPT-5.6 deployment.

OpenAI attributes the accuracy improvement to an internal evaluation of financial, medical, and legal prompts requiring factual detail. It reports that responses with at least one factual error were about 68% less common for the updated Sol than for GPT-5.5 Instant; it reports 62% fewer for Luna in the same setup. Those are vendor-reported results, not an independent benchmark, so they are useful as a description of the intended direction of the update but do not establish a comparable error rate for a particular production workload. The published example also evaluates answer focus in a weather-style dialogue, which is illustrative rather than a reproducible test protocol.

For practitioners, the practical split is straightforward. A team using standard paid ChatGPT can test whether the new slider and the more concise default behavior help its interactive research, writing, or planning workflows. A Codex, ChatGPT Work, or API team should not assume the post changes its deployed Sol behavior, and should continue its own regression checks before changing prompts or quality expectations. Separately, Free and Go ChatGPT users are moving to Luna as the default with a Think option for harder questions; OpenAI says text chats are unlimited subject to abuse safeguards, while other tools retain separate limits.