OpenAI tunes GPT-5.6 Sol for ChatGPT, while leaving Codex unchanged

OpenAI's August 6 product update changes GPT-5.6 Sol inside ChatGPT's standard Chat experience for Plus and Pro users. The company says the revised model is intended to answer more directly, use tighter formatting, adjust detail to the question, and offer corrections rather than merely agree. It also makes the same model serve both Instant responses and deeper reasoning, with a slider on web, mobile, and desktop for choosing how much thought an answer receives. This is a product-behavior update to an existing model family, not evidence of a separate API-model release.

The timing matters because GPT-5.6 Sol, Terra, and Luna reached general availability on July 9 across ChatGPT, Codex, and the API. OpenAI positioned Sol as the flagship tier, Terra as the lower-cost balanced tier, and Luna as the faster, lower-cost tier. The August announcement narrows its scope explicitly: the new Sol version is optimized for everyday ChatGPT chats and is available only in the Chat experience, while the Sol version used in Work and Codex is not changing. That distinction means a better or shorter answer in consumer ChatGPT should not be read as a change to an existing Codex workflow.

OpenAI's central quality claim is also deliberately bounded. It reports that, in its internal evaluation of factual financial, medical, and legal prompts, responses with at least one factual error were about 68% less common for the updated GPT-5.6 Sol than for GPT-5.5 Instant. The company attributes that result to better use of sources for questions involving dates, numbers, rules, and assumptions. Because this is an OpenAI internal evaluation rather than an independently reproduced benchmark, it is useful evidence of the intended direction of the update but not a universal accuracy guarantee for technical, scientific, or high-stakes use.

The change also sharpens the product split between paid and free ChatGPT. GPT-5.5 Instant remains the default fast model, while eligible paid users can select Medium, High, or Extra High reasoning with GPT-5.6 Sol; Pro is served by GPT-5.6 Sol Pro. Free and Go users do not receive Sol in standard ChatGPT conversations: GPT-5.6 Luna becomes their default, and the forthcoming Think control uses Luna rather than Sol. Existing plan limits still apply, and ChatGPT may fall back to GPT-5.4 Thinking mini after a Sol reasoning allowance is exhausted.

For developers and teams, the practical takeaway is to test the revised ChatGPT behavior where it will actually run instead of assuming a cross-product upgrade. The documentation continues to list Sol, Terra, and Luna in the OpenAI API and says the August Sol revision does not alter Work or Codex. Teams using ChatGPT for research, planning, writing, or decisions may find the slider useful for trading response speed against more extensive reasoning, but should retain task-specific evaluation, source checks, and fallback handling. The update improves a user-facing interaction surface; it does not remove normal model-access limits or turn provider-reported factuality results into an independent assurance.