Factor version soma-ai-ef-2026.10-dataset-v3.0.0, dated . Location-based grid averages, not renewable contracts. Estimates from published benchmarks.
A query of 1210 tokens to a mid-size AI model (class B, such as gpt-4o) served from US South Central - Texas emits about 0.043 g CO₂e, counting electricity, the share of server hardware, data-centre construction and amortised training; a million tokens emit 0.0312 kg CO₂e. A small model (class A) emits 0.013 g per query and a frontier model (class C) 0.16 g.
US South Central - Texas ranks 11 of 15 regions for class B, where 1 is the lowest carbon. Compared with EU North - Sweden, the same query here emits 4.4 times as much; compared with Asia Pacific - Singapore, 0.8 times as much. The grid, not the model, sets most of the difference.
| Model class | Electricity kg | Hardware kg | Construction kg | Training kg | Total kg CO₂e | g per query | Water L | mL per query |
|---|---|---|---|---|---|---|---|---|
| Class A (small) | 0.00817 | 0.00027 | 0.00014 | 0.00060 | 0.009 | 0.013 | 0.032 | 0.045 |
| Class B (mid-size) | 0.02675 | 0.00101 | 0.00046 | 0.00293 | 0.031 | 0.043 | 0.106 | 0.15 |
| Class C (frontier) | 0.09962 | 0.00388 | 0.00169 | 0.01284 | 0.118 | 0.16 | 0.393 | 0.54 |
Example: a company of 100 people where 80 use a class B assistant for 2,000 messages a month each sends about 2,323 million tokens a year. Served from US South Central - Texas, that is 72 kg CO₂e a year, against 17 kg in EU North - Sweden.
A 1210-token query to a class B model served from US South Central - Texas consumes about 0.15 mL of water, and a million tokens 0.106 litres. The figure combines the water evaporated on site for cooling with the water embedded in generating the grid electricity. Regions that Li et al. 2025 do not cover directly use an estimated on-site intensity, marked as such in the dataset.
Class B, total kg CO₂e per million tokens, lowest first: EU North - Sweden 0.00714, EU West - France 0.0082, EU North - Finland 0.0102, EU West - Belgium 0.014, EU West - UK (London) 0.0212, Netherlands 0.0235, EU West - Ireland 0.0245, US East - N. Virginia 0.0261, US West - Oregon 0.0275, EU Central - Germany 0.0303, US South Central - Texas 0.0312, US South - Southeast 0.0351, US Central - Iowa 0.0381, Asia Pacific - Japan 0.0395, Asia Pacific - Singapore 0.0401.
Electricity term: GPU energy per token from the ML.ENERGY Leaderboard v3, a facility overhead of 1.20, and grid carbon intensity from EPA eGRID 2023 and Ember 2023. Embodied hardware, data-centre construction and amortised training terms are documented on the methodology page. Water combines on-site cooling water (Li et al. 2025) with the water embedded in grid electricity (Reig et al. 2020). The modelled range of the models inside a class, around the class value, is −41% to +48% for class A, −39% to +24% for class B and −67% to +55% for class C.
Dataset: AI Inference Emission and Resource Factors for Corporate GHG Inventories, Llopis, G., SOMA AI, October 2026, CC BY 4.0, concept DOI 10.5281/zenodo.20443585, which resolves to the latest version (this version 10.5281/zenodo.23371447). Method: Accounting for Purchased AI Inference in Corporate GHG Inventories: A Token-Based Method, arXiv:2606.10660, under peer review.
Partners can fetch any row from the AI factors API.
When citing a single value, name the region and class, for example "0.0312 kg CO₂e per million tokens, class B, US South Central - Texas, SOMA AI factors October 2026", and link to app.somaai.earth/ai-factors/region/us-texas.