Arthur

Arthur provides monitoring and evaluation for machine-learning models and LLM applications. It tracks accuracy, drift and bias, and offers firewall-style protections against harmful outputs.

Its Arthur Engine is open source, so teams can run evaluations on their own infrastructure. Dashboards help data scientists and risk teams look at the same numbers.

It sits in the same space as Arize and Fiddler and tends to be chosen by financial and enterprise customers.