Launch Dates Reported for GPT-4.5 and GPT-5

OpenAI just dropped some eye-popping new numbers—and is reportedly gearing up for two major AI model launches in rapid succession.

The company now boasts 400 million weekly active users, marking a 33% jump in less than three months, alongside 2 million paying enterprise customers. (So much for claims about ChatGPT being on the decline.)

But the bigger headlines come from OpenAI’s rumored roadmap. The company plans to ship GPT-4.5 (codenamed “Orion”) as soon as this week, then follow it up with GPT-5 in late May. This pair of releases could fundamentally alter how we interact with OpenAI’s large language models—and maybe even change how businesses and knowledge workers everywhere approach their daily AI use.

To find out what’s really going on—and how to prepare—I talked to Marketing AI Institute founder and CEO Paul Roetzer on Episode 137 of The Artificial Intelligence Show.

What You Need to Know About GPT-4.5 and GPT-5

We know that GPT-4.5 is imminent. We also know that it’s OpenAI’s last model that won’t be “chain of thought."Roetzer, for one, doesn’t expect an out-of-this-world leap in capabilities when it drops.

“I'm not sure that 4.5 is going to be some amazing leap forward,” he says.

Instead, it’s more likely to lay the groundwork for the real breakthrough: GPT-5.

Arriving as early as late May, GPT-5 is where the real magic may happen. Sam Altman has previously said that the company wants to unify its GPT and o-series reasoning models into a single more powerful system capable of advanced reasoning, aka GPT-5.

Previously, OpenAI has floated five levels of AI that it intended to build, says Roetzer. These are:

• Level 1: Chatbots
• Level 2: Reasoners (human-level problem solving)
• Level 3: Agents (AI that can take actions)
• Level 4: Innovators (AI that can assist with groundbreaking inventions)
• Level 5: Organizations (AI that can effectively do the work of an entire enterprise)

We already have Chatbots (Level 1) in the form of the GPT models. We already have Reasoners (Level 2) in the form of the o-series models. Roetzer says that right now we’re in the midst of advancing Reasoners while also exploring Agents (Level 3).

Soon, all three levels—and levels beyond—could be unified in a single platform.

OpenAI may be pulling out all the stops on GPT-5 because it’s facing some of its stiffest competition to date. xAI’s Grok 3 model just dropped with advanced reasoning capabilities. Anthropic also just released Claude 3.7 Sonnet. 

This is the new normal, says Roetzer. For a long time, OpenAI was the clear leader in state-of-the-art models, with GPT-4 easily outpacing the competition. But now? 

“Everybody all of a sudden caught up.”

Do You Really Need to Try Every New Model?

So what do you actually do about this? With so many new releases coming so quickly, it’s easy to feel like you’re always behind. 

Roetzer suggests that you probably don’t need to chase every single new OpenAI release—or competitor model that arrives.

“As these models consolidate in their capabilities, people are going to switch models less and less,” he says. “You’re going to say: ChatGPT’s just good enough for me. Or I’ve got Gemini built into Workspace and it’s good enough.”

Even if the AI model you choose lags behind the truly state-of-the-art offerings by a couple months, most people likely won’t care, because what they use is good enough.

“My general advice would be: Just work with, and get good with, one of the models,” says Roetzer. “Assume it’s going to be good enough for what you need to do.”