Every model, both charts

EU-hosted, regulation-aligned.

EU AI Act aligned, French low-carbon grid, GDPR-native data handling.

Long-context reasoning with a published safety policy.

Publishes responsible scaling policy and supports binding safety regulation.

Huge context window, strong analysis.

Water reporting aggregated at fleet level, not per site.

Frontier generalist with deep reasoning.

Strong benchmarks; limited disclosure of per-query energy and water intensity.

Open-weight frontier you can self-host.

Open weights allow local inference; parent company scores poorly on water siting.

Answer engine that routes queries across multiple frontier models with citations.

Estimated scores pending confirmation. Multi-model aggregator, so impact varies with the routed provider.

Office-embedded assistant running on OpenAI frontier models.

Estimated scores pending confirmation. Inherits OpenAI model behaviour; enterprise tenants get no-training commitments.

Strong multilingual generalist.

Coal-heavy regional grid exposure across several compute regions.

Extremely cheap reasoning at scale.

Open weights, but opaque data provenance and no environmental disclosure.

Fast, unfiltered, realtime-leaning.

Memphis cluster ran unpermitted methane turbines; minimal transparency reporting.

Fully open data, weights, and training code.

Only model in the index publishing its full training corpus and compute footprint.

Balanced daily driver for code and prose.

Best blended performance-to-ethics ratio in the current index.

High-throughput multimodal at low cost.

Operator reports carbon-matched grid purchasing; evaporative cooling in several regions.

Fast, cheap frontier-adjacent workhorse.

Lower inference footprint per token than the flagship tier.