Multi-Platform, Multi-Model Generative AI Sustainability Cost Estimator
Estimate Calculations for Tokens, GHG, Water, and Material Consumption / 400 tokens
Hello!
[2025-Jul] This is a branch by Karol See to extend JV and Debbie's initial cost estimator by including sustainability metrics in Gen AI use cases.
Notes and thoughts:
Mistral's Sustainability costs found here, only shows Mistral Large 2 computation which is used for their general purpose chat service. While their API token costs is under Mistral Large 3.
Since this is considering public chat utilization, keep in mind that for threaded conversations previous context is being sent back to the API constantly
While this is just an estimate transition from word count to token count, computation is based on 1 token = 0.75 words or 1 word = 1.3333 tokens as defined below:
[2024-Mar] Thanks for dropping by and we hope you found this useful.
This is an early beta and will be updated regularly to cover more platforms, more models, self-hosted/instance-hr billing (e.g., SageMaker or Hugging Face) instead of just SaaS-per-token costing, per-character inference estimation, distinct dev & production costing, and more. Stay tuned!
This project is on GitHub: Gen AI Cost Estimator. Please feel free to send suggestions or bug reports there, we appreciate it!