Ecologits
🌱 EcoLogits tracks the energy consumption and environmental footprint of using generative AI models through APIs.
Last verified:
What is Ecologits?
EcoLogits is a suite of open-source tools for estimating the environmental footprint of generative AI models at inference. Based on life-cycle assessment (LCA) principles as defined by ISO 14044, the project raises awareness about the direct environmental impacts of AI while empowering developers and organizations to build more sustainable AI-powered applications. The tool tracks energy consumption and environmental impacts of using generative AI models through APIs without requiring code changes to existing requests.
The main components include: (1) EcoLogits Calculator, a user-friendly web tool for estimating environmental footprint with just a few clicks; (2) EcoLogits Python library, which tracks environmental impacts of AI requests made through official client SDKs; and (3) EcoLogits API, an API service enabling integrations of impact assessment methodology into applications. The library supports major LLM providers including OpenAI, Anthropic, Mistral AI, Cohere, Google Generative AI, and Hugging Face Inference Endpoints.
EcoLogits reports environmental impacts across five criteria: Energy (final electricity consumption in kWh), Global Warming Potential/GWP (GHG/carbon emissions in kgCO2eq), Abiotic Depletion Potential for Elements/ADPe (depletion of minerals and metals in kgSbeq), Primary Energy/PE (total energy from primary sources in MJ), and Water Consumption Footprint/WCF (water consumption in liters). Impacts are reported for two life cycle phases: Usage (energy consumption during model execution) and Embodied (resource extraction, manufacturing, and transportation).
The tool is designed for developers building AI applications, sustainability managers in organizations, researchers studying AI environmental impacts, and anyone wanting to understand and reduce the environmental footprint of their generative AI usage. It is actively maintained by the GenAI Impact non-profit and is part of the CodeCarbon non-profit ecosystem.
Ecologits pricing
Pricing model: Freemium
EcoLogits is completely free and open source under the Mozilla Public License Version 2.0 (MPL-2.0). The Python library can be installed via pip (pip install ecologits), the web Calculator is freely accessible, and the API service at api.ecologits.ai/v1beta is available for integration. There are no paid tiers or subscription plans. The project is maintained by the GenAI Impact non-profit and supported by sponsors including Data For Good and Boavizta.
Ecologits pros
- Open source with transparent code, methodology, and data
- Seamless integration without changing existing API request code
- Supports major LLM providers: OpenAI, Anthropic, Mistral AI, Cohere, Google, Hugging Face
- Tracks five environmental impact criteria comprehensively
- Reports impacts per life cycle phase (Usage and Embodied)
- Based on ISO 14044 Life Cycle Assessment methodology
- Uses bottom-up modeling approach for precise impact allocation
- Provides high-confidence approximation intervals for estimates
- Includes user-friendly web calculator for quick estimates
- Python library with official client SDK integration
- API service available for custom application integrations
- Actively maintained by GenAI Impact non-profit organization
- MPL-2.0 license allows flexible use and modification
- Community-driven development with collaborative improvements
- Tracks inference impacts without covering training overhead
- Returns ImpactsOutput object with impacts, warnings, and errors
- Supports Python 3.10+ with Pydantic and Wrapt libraries
Ecologits cons
- Does not cover end-of-life phase due to e-waste recycling data limitations
- Excludes networking impacts from environmental assessment
- Excludes end-used device impacts from assessment
- Excludes model training impacts, only tracks inference
- Relies on provider data which may be incomplete or proprietary
- Estimates are approximations due to lack of open transparent data from cloud providers
- Some model architectures may trigger warnings if not released publicly
- Requires Python 3.10+ which limits older Python version compatibility
- Multimodal models may have limited architecture information available
- Methodology requires providers to share information for customization
Frequently asked questions about Ecologits
What is EcoLogits and what does it do?
EcoLogits is a suite of open-source tools for estimating the environmental footprint of generative AI models at inference. It tracks energy consumption and environmental impacts of using generative AI models through APIs, supporting major LLM providers like OpenAI, Anthropic, and Mistral AI. The tool reports impacts across five criteria (Energy, GWP, ADPe, PE, WCF) and two life cycle phases (Usage and Embodied) based on ISO 14044 Life Cycle Assessment methodology.
How do I install EcoLogits?
To install EcoLogits, use pip install ecologits. You can select specific providers during installation, for example pip install ecologits[openai] for OpenAI integration. It requires Python 3.10+ and relies on Pydantic for data modeling and Wrapt for function patching. It is also possible to install EcoLogits without any provider if you plan to add them later.
Which AI providers does EcoLogits support?
EcoLogits supports major LLM providers including OpenAI, Anthropic, Mistral AI, Cohere, Google Generative AI (google-generativeai), and Hugging Face Inference Endpoints (huggingface-hub). The library uses function patching to work with official client SDKs from these providers without requiring code changes to your API requests.
What environmental impact metrics does EcoLogits track?
EcoLogits tracks five environmental criteria: Energy (final electricity consumption in kWh), Global Warming Potential/GWP (carbon emissions in kgCO2eq), Abiotic Depletion Potential for Elements/ADPe (mineral and metal depletion in kgSbeq), Primary Energy/PE (total energy from primary sources in MJ), and Water Consumption Footprint/WCF (water consumed in liters). These are reported for both Usage phase (energy during execution) and Embodied phase (resource extraction, manufacturing, transportation).
Do I need to change my existing code to use EcoLogits?
No, you do not need to change your code when making API requests. You only need to initialize EcoLogits with EcoLogits.init(providers=['openai']) before creating your client. After initialization, your existing API calls automatically return an ImpactsOutput object containing environmental impacts, warnings, and errors alongside the normal response.
What is the EcoLogits Calculator?
The EcoLogits Calculator is a user-friendly web tool for estimating the environmental footprint of AI models with just a few clicks. It provides a no-code interface for users who want to quickly estimate impacts without integrating the Python library or API into their applications.
What methodology does EcoLogits use for impact assessment?
EcoLogits employs the Life Cycle Assessment (LCA) methodology as defined by ISO 14044. It uses a bottom-up modeling approach to assess and aggregate environmental impacts of individual service components. The methodology focuses on GenAI inference tasks and is grounded in transparency and reproducibility, utilizing open market and technical data. It computes high-confidence approximation intervals for estimates.
What is included in the ImpactsOutput object?
The ImpactsOutput object returned for each request gathers environmental impacts and potential warnings and errors. It contains total estimated energy consumption in kWh, total estimated GHG emissions in kgCO2eq, and impacts broken down by life cycle phase (Usage and Embodied) and by criteria (Energy, GWP, ADPe, PE, WCF). It also includes has_warnings, warnings array, has_errors, and errors array for handling issues.
What are the limitations of EcoLogits?
EcoLogits does not cover the end-of-life phase due to data limitations on e-waste recycling. It also excludes networking impacts and end-used device impacts from its assessment, focusing only on hosting and running model inferences. Estimates are approximations due to lack of open transparent data from AI/cloud providers and hardware manufacturers. Some proprietary model architectures may have limited information available.
Who maintains EcoLogits and what is its license?
EcoLogits was created and is actively maintained by the GenAI Impact non-profit organization. The project is part of the CodeCarbon non-profit and is supported by Data For Good and Boavizta. The Python library is licensed under Mozilla Public License Version 2.0 (MPL-2.0), and all methodologies are licensed under CC BY-SA 4.0, requiring proper citation of EcoLogits and its authors.