Trustible Logo

Model Transparency Ratings

Trustible's Model Transparency Ratings are a guide to help AI deployers, AI governance professionals, and the general public make more informed decisions with respect to model selection. Trustible has evaluated the transparency and risk disclosures of popular AI models. The ratings are based on regulatory requirements, like the forthcoming EU AI Act, and Trustible's AI Risk Taxonomy. Our ratings are based on publicly available documentation from each developer and are linked on individual model pages, alongside explanations behind our methodology. If you have any feedback or questions on our Model Transparency Ratings, please email us at contact@trustible.ai

MODEL

Open Weight

Open Data

Systemic Risk

Overall Transparency

Data Transparency

Model Transparency

Systemic Risk Evaluation Transparency

Arctic High High Medium Not Applicable
AYA High High Medium Not Applicable
BERT Family High Medium High Not Applicable
Claude 3 Family Medium Low Low High
Command R Medium Low Low Not Applicable
DBRX Medium Low Medium Not Applicable
Falcon Family Medium Medium Low Not Applicable
Gemini 1.5 Family Medium Low Medium Medium
Gemini 1 Family Medium Low Medium High
GPT-3.5 Medium Low Medium Not Applicable
GPT-4 Medium Low Medium Medium
Grok Low Low Low Not Applicable
Llama-2 High Medium High Not Applicable
Llama-3 Family Medium Low Medium Not Applicable
Mistral-7B + Mixtral-8x7B Medium Low Medium Not Applicable
Mistral-Large Low Low Low Low
MPT Family High High High Not Applicable
OLMO Family High High High Not Applicable
Phi-2 Medium Medium Medium Not Applicable
Phi-3 Family High Medium High Not Applicable
Titan Low Low Medium Not Applicable
Vicuna Family Medium Medium Medium Not Applicable
YI Family Medium Medium High Not Applicable