GPT-5.6 Sol's ChatGPT update is a product-layer change, not an API revision

OpenAI updated GPT-5.6 Sol in ChatGPT for Plus and Pro users on August 6, saying the chat experience now gives more focused answers, improves factual reliability, and lets users select how much thought to apply with a slider. It also says the same Sol model now powers both Instant responses and deeper reasoning in ChatGPT. The immediate change is therefore behavior and control inside the consumer chat product, rather than a newly named model endpoint.

The update follows the GPT-5.6 family's July rollout. GitHub's changelog independently documented Sol, Terra, and Luna as distinct choices in Copilot, describing Sol as the highest-reasoning tier for large codebases and long-running agentic work. That context matters because OpenAI's August announcement changes how Sol behaves in ChatGPT while the wider product ecosystem still presents the family as a tiered model selection problem.

OpenAI says internal evaluations of financial, medical, and legal prompts found at least one factual error about 68% less often with GPT-5.6 Sol than with GPT-5.5 Instant. The claim is specifically an internal result and the examples are designed to illustrate answer style, so it should not be read as an independently replicated general benchmark. The more defensible takeaway is that OpenAI is targeting source use, assumptions, numbers, and dates as a product-quality problem in ChatGPT.

The important boundary for technical users is explicit: OpenAI says the updated Sol version is available only in the Chat experience, and that the GPT-5.6 Sol version powering Work and Codex is not changing. The API model page continues to identify GPT-5.6 Sol as the destination of the gpt-5.6 alias, with its existing context window, pricing, endpoints, and tool support. The announcement does not describe an API contract, pricing, or Codex behavior change.

Teams should therefore test the update where it exists, in ChatGPT, rather than assume a prompt, agent, or evaluation pipeline using the API has changed underneath them. The new reasoning slider may improve interactive task calibration for paid ChatGPT users, while free users are being moved toward Luna with a separate Think control. For developers who need stable production behavior, the documented API configuration remains the relevant source of truth; the ChatGPT announcement is a useful product signal, not a substitute for deployment validation.