OpenAI says 80 to 90 percent of its research already targets 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 because that’s where “most of the value” comes from. Improvements within a generation, like going from GPT 5.1 to 5.2, come from specialized training data but are intentionally short-term bets.

The real leap comes with a new model generation, where “everything else just works a lot better,” and after each jump, the company has to relearn where to invest for quick wins. These incremental updates “are seen as extremely shortsighted” within the company, but they help OpenAI iterate and learn faster today, although it’s not the right long-term strategy, Power says. Power sees the main challenge with today’s AI assistants not in model quality but in onboarding. Most ChatGPT users don’t know what they can do with AI, he says. Future models should be better at surfacing what’s possible and anticipating what users need. GPT-4 required careful prompting. GPT-5 is easier to use but still demands a lot of feedback. GPT-6 already works more like a capable colleague you can just hand a goal to, Power says. Follow The Decoder for AI news, background stories and expert analyses.