AI is advancing fast and countries must better use today’scapabilities to close the gap January 2026 A capability overhang is growing as innovation outpacesadoption The history of general purpose technologies – from steam engines to semiconductors – suggests thatthe biggest economic gains come not from invention alone, but from turning new capabilities intoscaled, everyday use. AI is advancing at unprecedented speed: the length of tasks it can perform hasbeen doubling roughly every seven months, and progress does not appear to be slowing.In 2022,frontier models could reliably complete tasks that take human experts about 1 minute to complete;today, they can complete tasks that take human experts more than 30 minutes.1However, increasedcapabilities alone do not lead to productivity benefits or economic impact. To capture these benefits,countries, companies, and people need to fully use AI tools in real workflows to solve real problems. As AI capabilities have improved, we see a widening “capability overhang,” defined as the gap betweenwhat AI tools can do and how typical users are using them. This overhang increases as capabilitiesprogress beyond simple chat to more sophisticated, multi-step workflows. The capability overhang isalso associated with measurable productivity impact, as workers who use more advanced capabilitiesreport higher time savings.2Across an economy, these gains compound, freeing people to tackleharder tasks, build new products and services, and accelerate innovation that raises growth and livingstandards. At its core, this overhang reflects gaps not just in access but in agency. Access is the admission ticket:without it, people, organizations, and institutions can’t participate fully in the AI era. Agency is whatturns access into impact: the ability and incentive to deploy AI meaningfully in real work. When accessand agency align, more people can participate not just as users of AI, but as active drivers of the growthit enables. Addressing this capability overhang is central to people and countries realizing AI’s full economic andsocial potential. It’s also central to OpenAI’s mission to ensure that artificial general intelligence benefitsall of humanity, which means doing our part to close the gap between access and impact by expandingboth the availability of AI tools and people’s ability to deploy them effectively in real work. By enablingmorepeople,organizations,and institutions to move from basic use to deeper,more capable workflows, we can help AI scale human ingenuity, drive productivity and growth, and create opportunityfor people everywhere, not just a few early adopters. ChatGPTusage data shows how large this overhang has become. To estimate how deeply AIcapabilities are being used, we measure the amount of effort the model uses to respond to userqueries, reflecting usage depth. When users ask more difficult and complex questions, the modeltypically spends more effort reasoning to produce a better answer. Counting reasoning tokens is a wayof measuring this effort, and throughout this report we refer to this metric as “thinking capabilities.” Taskand tool usage are also important measures of capability adoption, reflecting both breadth and depth ofAI capabilities. The typical power user uses 7x more thinking capabilities than the typical user.3We alsoobserve a clear country-level gap. Across more than 70 countries with the highest ChatGPT users,leadingcountries use 3×more thinking capabilities per person than users in laggingcountries.Major economies like the United States and India lead in total users,and smallerhigh-incomecountries like Singapore and the Netherlands lead in population penetration.Still,advanced AI adoption is not confined to large or high-income countries, with nations like Vietnam andPakistan ranking among the world’s top users of agentic tools. Leading countries also use AI moredeeply, with more than 2x higher per person usage of key tasks and advanced tools like data analysis,Apps (formerly Connectors), Codex and more. This capability overhang will not disappear on its own and may even widen as some countries lead inapplying frontier AI capabilities and others fall behind. AI can drive broad-based growth, but only iffrontier capabilities are applied to help solve economically valuable problems. AI is for the IntelligenceAge what electricity was for the later Industrial Age and the internet was for the Information Age, andcountries that seize the opportunity will be ahead competitively, economically, and socially. Launched in 2025,OpenAI for Countriesis our initiative to help governments put AI and its benefits intothe hands of people around the world, aligned with local needs, institutions, and priorities. Now underthe leadership of former UK Chancellor of the Exchequer George Osborne, we are expanding this workwith new initiatives focused on education, health, AI skills training and certifications, cybersecurity,disaster relief, and startup accele