OpenAI’s Research Focus Is Already on GPT-7 and Beyond

Boris Power, OpenAI’s Head of Applied Research, says 80 to 90 percent of the company’s research goes toward GPT-7, GPT-8, and beyond. The reason, according to Power, is that this is where “most of the value” comes from. For readers following AI product updates, the key takeaway is that today’s visible improvements may represent only a smaller slice of OpenAI’s overall research effort.

Power describes improvements within a single generation, such as moving from GPT-5.1 to GPT-5.2, as driven by specialized training data. These updates are intentionally short-term bets and are viewed inside the company as “extremely shortsighted,” according to the article. Still, they help OpenAI iterate and learn faster in the present, even if they are not described as the long-term strategy.

Power does not frame model quality as the main challenge for today’s AI assistants. Instead, he says most ChatGPT users do not know what they can do with AI. That shifts the practical focus from raw capability alone to helping people understand, discover, and apply those capabilities more effectively.

Power says future models should be better at showing users what they can do and anticipating what users need. The article contrasts GPT-4, which required careful prompting, with GPT-5, described as easier to use but still requiring a lot of feedback. GPT-6 is described by Power as working more like a capable colleague that can be given a goal.