Microsoft AI Economy Institute estimates, republished by Our World in Data with ISO3 codes. 147 countries, all three periods. IMPORTANT: values are modelled, and low-data countries are imputed in regional blocks — 12 West African countries share exactly 10.1% and the four Guianas share exactly 10.3%. Rank movements inside those blocks are artefacts of the model, not independent national trends.
CSET estimates. 119 countries appear across 2016–2025, fewer in any single year. Absolute totals, not population-adjusted — small economies rank low by construction.
Epoch AI. Only ~35 countries have ever produced a notable model, and the United States and China dominate. Most of the map is legitimately no-data. Models credited to several countries count once for each.
190 countries appear across 2016–2024, fewer in any single year — the header above counts the latest year only. Best coverage in the project. Includes Kosovo, which OWID codes as OWID_KOS and which has no ISO numeric — it appears in tables and rankings but cannot be picked on the globe.
| Model | Organization | Domain | Training compute | Published |
|---|---|---|---|---|
| Aya | Cohere Labs (formerly Cohere for AI),Brown University,Cohere,Carnegie Mellon University (CMU),Massachusetts Institute of Technology (MIT) | Language | Not disclosed | 2024-02-12 |
| Cohere Embed | Cohere | Language | Not disclosed | 2023-11-02 |
| BlenderBot 3 | McGill University,Meta AI,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms) | Language | 4.3e23 | 2022-08-10 |
| EMDR | Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),McGill University,DeepMind | Language | 1.9e21 | 2021-06-09 |
| TaLK Convolution | Carleton University | Language | 2.7e19 | 2020-02-08 |
| HRA | Maluuba,Microsoft | Games | Not disclosed | 2017-06-13 |
| Prototypical networks | University of Toronto,Twitter | Vision | Not disclosed | 2017-03-15 |
| DeepStack | University of Alberta,Charles University,Czech Technical University | Games | 1.4e19 | 2017-01-06 |
| Gated HORNN (3rd order) | York University | Language | Not disclosed | 2016-04-30 |
| SC-NLM | University of Toronto | MultimodalVisionLanguage | Not disclosed | 2014-11-10 |
| GANs | University of Montreal / Université de Montréal | Image generation | 5.2e17 | 2014-06-10 |
| GRUs | University of Montreal / Université de Montréal,Jacobs University,University of Maine | Language | Not disclosed | 2014-06-03 |
Compute bars are log-scaled across the models shown, in FLOP. Many organisations do not disclose it.
#7
South Korea
10,265
#8
Italy
9,123
#9
Canada
This country
#10
France
7,681
#11
Australia
7,605