Editorial illustration for OpenAI Focuses 80-90% of Research on GPT-7 and Future Models
OpenAI Devotes 80-90% Research to GPT-7 and Beyond
OpenAI Focuses 80-90% of Research on GPT-7 and Future Models
Boris Power, OpenAI's Head of Applied Research, put a number on something the company rarely quantifies out loud: 80 to 90 percent of its research effort is already aimed at GPT-7, GPT-8, and whatever comes after. Not the model shipping next quarter. The one several generations out.
That's a strange place to point most of your resources if you're a company that just released GPT-5 and is still fielding complaints about it. But Power's logic, laid out in recent comments on OpenAI's research priorities, treats point releases like 5.1 to 5.2 as short-term patchwork, useful for learning fast, not for building the future. The real gains, he argues, show up when a new generation lands and "everything else just works a lot better," forcing the company to relearn where the quick wins even are.
He also flags a problem that has nothing to do with model architecture: most people using ChatGPT still don't know what it can actually do for them. Below, Power explains how that gap between model capability and user awareness has changed from GPT-4 to GPT-6, and why he considers OpenAI's own incremental updates "extremely shortsighted."
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.
Why this matters
Power's numbers tell developers and founders something worth sitting with: if 80 to 90 percent of OpenAI's research firepower is aimed at GPT-7 and beyond, the point releases we obsess over, 5.1 to 5.2, are basically table scraps by design. That's a useful recalibration for anyone building product roadmaps around incremental OpenAI updates. Treat them as maintenance, not signal.
The real architectural shifts, the ones that change what's economically viable to build, arrive with generational jumps, and those are further out and less frequent than the release cadence suggests. For researchers, there's a candid admission buried here too: OpenAI itself says it has to relearn where the quick wins are after every big jump, which means even insiders are guessing at optimization paths short-term. That's a rare bit of honesty from a company usually selling certainty.
Worth watching: whether "extremely shortsighted" internal framing changes how OpenAI communicates version updates externally, or whether we keep getting incremental releases dressed up as bigger news than Power's own comments suggest they are.
Related Reading
Browse all OpenAI news →
Research & Benchmarks Microsoft AI Chief: "If It Isn't Safe We Shouldn't Build It
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives AI access drops correct answers by two-thirds, study shows
From Archives Nvidia’s SoL-Pi halves coding agent token use with evidence-preserving reducer
From Archives OpenAI Pauses AI Training After Security Probes Reveal Vulnerabilities
Open Source
From Archives OpenAI Halts Training of Top Models Amid Concerning Behavior Reports
Business & Startups
Common Questions Answered
What percentage of OpenAI's research effort is focused on GPT-7 and future models?
According to Boris Power, OpenAI's Head of Applied Research, 80 to 90 percent of the company's research effort is already aimed at GPT-7, GPT-8, and whatever comes after. This allocation reflects OpenAI's strategic focus on long-term development rather than immediate product releases.
Why is OpenAI allocating most of its research resources to models several generations ahead?
Boris Power explains that OpenAI concentrates its research on future models because that's where 'most of the value' comes from. This forward-looking approach prioritizes architectural shifts and breakthrough capabilities over incremental improvements to current models.
How should developers view point releases like GPT-5.1 and GPT-5.2 given OpenAI's research priorities?
According to the article, developers should treat incremental point releases as maintenance updates rather than signals of major innovation. Since 80-90 percent of OpenAI's research targets future generations, these minor updates represent limited resources compared to the effort invested in next-generation models.
What does OpenAI's research allocation tell us about building product roadmaps?
The article suggests that companies building around incremental OpenAI updates should recalibrate their expectations, as most of OpenAI's research firepower is directed toward future models rather than near-term releases. Real architectural shifts and economically viable changes arrive with major new model generations, not point releases.
Further Reading
- Research | OpenAI - OpenAI
- Research Index - OpenAI
- OpenAI Research | Publication - OpenAI
- Deep Research System Card - OpenAI
- Boris Power - X
