Carbon footprint of small AI models per query and per token

One query of 1210 tokens to a small model (class A, such as gpt-4o-mini, gpt-4.1-mini, gpt-4.1-nano) emits about 0.016 g CO₂e on the global-average grid, and a million tokens emit 0.0115 kg CO₂e. Served from EU North - Sweden the same query emits 0.0025 g, and from Asia Pacific - Singapore 0.016 g.

For comparison on the same grid, a class B (mid-size) query emits 0.053 g and a class C (frontier) query emits 0.2 g. Class A uses 0.021769339805825245 kWh of facility electricity per million tokens, and its uncertainty band around the class factor is about 48%.

EU North - Sweden gives the lowest total for class A, 0.00185 kg CO₂e per million tokens and Asia Pacific - Singapore the highest at 0.0119 kg, a 6.4 times spread. The spread on electricity alone is 13 times.

A 1210-token query to a class A model consumes 0.045 mL of water in US South Central - Texas, the lowest-water region with a figure, and 0.29 mL in US West - Oregon, the highest. 14 of 18 regions have a water figure.

On the global-average grid, electricity is 91.2% of the class A total, hardware 2.3%, data-centre construction 1.2% and amortised training 5.2%. In EU North - Sweden electricity falls to 45.3% and in Asia Pacific - Singapore it rises to 91.5%. The conservative high case adds 61% to the class total.

The dataset's mapping table assigns 117 model name patterns from 14 providers to class A, and a star stands for any text. 14 of the 59 mapping rows for this class are flagged as reasoning models by default.

• OpenAI gpt-4o-mini, gpt-4.1-mini, gpt-4.1-nano, gpt-5-mini, gpt-5-nano, o1-mini, o3-mini, o4-mini, gpt-3.5-turbo, gpt-3.5, gpt-5*-mini, gpt-5*-nano, gpt-5*-terra, gpt-6*-terra, gpt-5*-luna, gpt-6*-luna, gpt-5*-codex-mini, gpt-6*-codex-mini, codex-mini, gpt-6*-mini, gpt-6*-nano, gpt-oss-20b
• Google gemini-*-flash, gemini-*-flash-lite, gemma-*-2b, gemma-*-4b, gemma-*-9b, gemma-*-12b, gemini-1.5-flash, gemini-flash, gemini-*-nano, gemini-nano, gemma-*-e2b, gemma-*-e4b, gemma-*-270m, gemma-*-1b, gemma-*-3b, gemma-*-7b, gemma-*-8b, gemma-*-26b-a4b
• Alibaba qwen*-flash, qwen*-turbo, qwen3-8b, qwen3-14b, qwen3-30b-a3b, qwen*-35b-a3b, qwen*-0.5b, qwen*-1.5b, qwen*-1.8b, qwen*-3b, qwen*-4b, qwen*-7b, qwen*-8b, qwen*-14b
• Meta llama-3.1-8b, llama-3.2-1b, llama-3.2-3b, llama-3.2-11b*, llama3-8b, llama3-1-8b, llama-3-8b, llama3-2-1b, llama3-2-3b, llama3-2-11b, llama-guard, llama-prompt-guard, prompt-guard
• Microsoft phi-3-mini, phi-3-small, phi-3-medium, phi-3.5-mini, phi-4-mini, phi-4-multimodal, phi-3.5-moe, phi-3-moe, phi-4, phi-
• Mistral ministral-*, mixtral-8x7b, mistral-tiny, mistral-nemo, pixtral-12b, leanstral, shieldstral, classifier-, mistral-7b, open-mistral-7b
• Cohere command-r7b, command-light, command-nightly-light, north-mini, c4ai-aya, aya-expanse-8b, aya-23-8b
• Amazon nova-micro, nova-*-micro, nova-lite, nova-*-lite, titan-text-lite
• Anthropic claude-haiku-*, claude-3-haiku*, claude-3-5-haiku*, claude-*-haiku, claude-instant
• DeepSeek deepseek-r1-distill-llama-8b, deepseek-r1-distill-qwen-1.5b, deepseek-r1-distill-qwen-7b, deepseek-r1-distill-qwen-14b
• NVIDIA nemotron-nano, nvidia-nemotron-nano, nemotron-*-nano
• Z.ai (Zhipu) glm-*-flashx, glm-4*-flash, glm-4*-air
• xAI grok-*-mini, grok-mini
• Xiaomi mimo-*-distill-qwen-9b

• Class A (small)
• Class B (mid-size)
• Class C (frontier)

Figures per query are one plain exchange, before the Tier 2b reasoning uplift. 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 (SOMA AI Emission Factors (SAEF)), Llopis-Garcia, G., SOMA AI, 19 October 2026, CC BY 4.0, concept DOI 10.5281/zenodo.20443585, which resolves to the latest version (this version 10.5281/zenodo.23371447, earlier version 2 values at 10.5281/zenodo.22767475). 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, and the whole table is a CSV download.

The full table for every class and region is on the SOMA AI Emission Factors (SAEF) page.

• Carbon footprint of mid-size AI models (class B)
• Carbon footprint of frontier AI models (class C)
• How the factors are calculated
• How to account for AI usage in a Scope 3 inventory, and when CSRD applies