AI models in the top 100
How many of the 100 highest-rated language models come from each region.
Where the top 100 language models come from
LMArena leaderboard · snapshot 2026-09-13
Follows LMArena's published style-controlled ranking, refreshed by the publisher every few days. Model variants count as the leaderboard lists them.
The story
Not all AI models are created equal. Many organisations build AI systems; only a handful reach the technological frontier. Those models set the benchmarks and become the foundation the next generation of AI products, discoveries and applications is built on.
This indicator counts how many of the world's top 100 AI models come from each region. It does not measure overall AI activity; it measures leadership at the frontier. A higher share means a country is not just adopting AI but defining the technology others build upon. Today the United States dominates frontier model development, and Europe holds only a small share.
Why it matters
Frontier AI models pull far more than the technology sector along with them. They attract world-class researchers, draw private investment and seed the environments where startups, universities and established companies build new products and services. The countries that lead at the frontier also get the largest say in global technical standards and platforms.
As AI becomes a main driver of productivity across the economy, leadership in frontier models is an increasingly direct read on a country's broader capacity for technological innovation.
Policy context
Developing frontier AI takes more than excellent research. It takes advanced computing infrastructure, highly skilled talent, risk capital and companies that can scale a breakthrough. Europe's recent competitiveness initiatives name exactly these foundations as the condition for staying in the global AI race.
Europe has strong research institutions and a growing AI ecosystem. Whether more of tomorrow's leading models are developed here depends on the conditions for commercialising that research and scaling the companies behind it.
How this number is calculated
Read from the LMArena leaderboard's published data, with each model's origin mapped to a region by its publishing organisation. Shown as a dated snapshot, refreshed whenever the leaderboard publishes a new ranking, and holding the last reading in between.
Source & cadence
Related policy insights
No insight references this indicator yet; new analyses appear here automatically.
Browse insights