The news: The release of OpenAI’s long-awaited GPT-5—a frontier model the company originally expected to launch in summer 2024—hit turbulence almost immediately.

• Despite high expectations, early users reported the model felt sluggish and less capable than GPT-4o, labeling it “kinda mid.” It’s a surprising letdown for what was billed as a major leap forward.
• A viral “chart crime” from the livestreamed demo, where a shorter score was represented by a taller bar, drew widespread ridicule. OpenAI CEO Sam Altman later called it a “mega chart screwup” on X.

Altman pledged to address early problems and double rate limits to encourage more experimentation with GPT-5. He added that OpenAI may restore GPT-4o access for Plus users, showing a backtrack on OpenAI’s hope for the new model to serve as the be-all-end-all.

Under the hood: GPT-5’s biggest pitch is its dynamic “thinking,” but the tech behind that stumbled. The autoswitcher, responsible for independently deciding how hard to “think” about responses and how much compute to use, was out of commission at launch, leading to inconsistent performance.

In addition, rival products may still outperform it. Users report that Anthropic’s Claude Opus 4.1 and Alibaba’s upgraded Qwen3 work better than GPT-5 in real-world use cases, such as solving math problems and coding tasks.

Why it matters: First impressions are sticky, and rolling out a flagship AI model with broken features and visible presentation errors risks eroding user trust.

The autoswitcher misstep undermined one of GPT-5’s key selling points, while the chart blunder reinforced perceptions of carelessness. For a company with 700 million weekly active users, even a small drop in perceived quality can have outsized reputational consequences.

Faster releases may keep OpenAI competitive, but poorly staged unveilings hand narrative control to competitors. As Apple’s Steve Jobs demonstrated in his turnaround era, meticulous launch preparation can turn a product debut into a cultural event rather than a public troubleshooting session.

Our take: GPT-5 itself will likely improve over time, and its technical advances shouldn’t be dismissed, but the lesson is clear—being first out of the gate is only valuable if the presentation matches the product’s ambition.

Marketing and communications remain stubbornly human domains for now. If AI could fully replace them, OpenAI’s own product announcements would run like clockwork. Instead, the debut of one of the world’s most advanced AI models was labeled an avoidable public relations headache, showing that even cutting-edge technology may be remembered less for what it can do and more for how it was introduced.

This content is part of EMARKETER’s subscription Briefings, where we pair daily updates with data and analysis from forecasts and research reports. Our Briefings prepare you to start your day informed, to provide critical insights in an important meeting, and to understand the context of what’s happening in your industry. Non-clients can click here to get a demo of our full platform and coverage.