OpenAI shipped GPT-5.6 to ChatGPT — the model that answers questions about your business just changed on the biggest AI engine

What OpenAI shipped

Per TechCrunch's July 9, 2026 report, corroborated by CNBC and Engadget, OpenAI made its GPT-5.6 family generally available across ChatGPT, Codex, and the OpenAI API. The specifics that matter:

• Three tiers, one family. GPT-5.6 ships as Sol (the flagship “workhorse,” which OpenAI calls its best coding model yet), Terra (the intermediate option), and Luna (the budget option). API pricing runs $5/$30 per million input/output tokens for Sol, $2.50/$15 for Terra, and $1/$6 for Luna.
• It rolled out to ChatGPT, not just the API. TechCrunch reports availability “across ChatGPT, Codex, and the OpenAI API” — so the model consumer users talk to when they ask ChatGPT a question now includes GPT-5.6, with the global rollout continuing over the following day.
• It cleared a government cybersecurity review first. GPT-5.6 was previewed to a “small group of trusted partners” on June 26 at the U.S. government's request under an early-June AI cybersecurity order that asks companies to submit their most powerful models 30 days before public release. Per CNBC, the Department of Commerce's Center for AI Standards and Innovation ran additional testing before clearing the wider July 9 launch.
• OpenAI calls it its strongest cybersecurity model yet — the exact capability that triggered the review — while positioning Sol primarily as a coding and agentic workhorse.

Why a model swap on ChatGPT resets your AI-visibility numbers

An answer engine's recommendations are a property of the specific model running underneath it. Change the model and you change the answers — which businesses get named, in what order, with what accuracy. That is not a subtle effect: in our own run-to-run study, only 31% of recommended brands survived every repeat of the same question on the same engine, before any model even changed. A frontier-model release is a step change stacked on top of that everyday flicker. Any screenshot, agency report, or dashboard column that measured “ChatGPT” before July 9 describes model behavior your buyers may already have moved past.

This is the same reset we flagged when Anthropic swapped Claude's default to Sonnet 5 on June 30 — except ChatGPT is the higher-stakes surface. Previsible's data shows ChatGPT drives 92% of trackable standalone-LLM referral traffic; it is the AI engine the largest share of your customers actually use. A model change there moves more of your real-world visibility than a change anywhere else.

What this means for a local business specifically

If you run a med spa, a law firm, or a home-services company, the practical exposure is simple: the assistant a customer opens to ask “who's the best {your category} near me” is now, at least in part, a different model than it was last week — and the engines already disagree sharply about who to name. In our 1,120-answer comparison, ChatGPT recommended the audited business in just 16% of answers, less than half Claude's 38%. You don't get told when the model behind that number changes; you only find out if you re-measure. And because ChatGPT answers in-place, an AI that stops naming you produces no lost click in any analytics tool — just a booking that quietly goes to whoever it named instead.

The takeaway

Treat July 9 as a reset line for your ChatGPT numbers, the way June 30 was for Claude. Engines change the model underneath you with no notice to the businesses being recommended — so the only defense is to re-measure after each swap rather than trust a stale report. A free 60-second check samples ChatGPT and Perplexity (the $79 audit adds Gemini and Claude, 25 questions sampled three times each) and reports how often each one names your business right now, and if a tool is tracking this for you, confirm it has re-sampled since GPT-5.6 shipped (here's how the monitoring tools compare).