Patronus AI

Patronus AI's Etsy case describes using Judge-Image confidence scores and reasoning alongside its Experiments framework to compare multimodal captioning systems and mitigate product-image caption hallucinations.

Patronus reports numerous Fortune 500 and leading AI customers, millions of evaluation requests and hundreds of thousands of detected hallucinations and failures.

Patronus reports numerous Fortune 500 and leading AI customers, millions of evaluation requests and hundreds of thousands of detected hallucinations and failures.

Patronus introduces GLIDER, a 3.8B evaluation model generating reasoning and text highlights with configurable binary or Likert-scale scoring, released on Hugging Face. The company reports 91% agreement with human judgment in its evaluation; that figure is not independently reproduced here.

Patronus develops LLM evaluation, hallucination detection, safety benchmarks and research-first evaluation models such as Lynx and FinanceBench.

Patronus develops LLM evaluation, hallucination detection, safety benchmarks and research-first evaluation models such as Lynx and FinanceBench.

Patronus announced a $50 million Series B led by Greenfield Partners in June 2026 with participation from Lightspeed, Notable Capital, Datadog and Samsung; the company described planned research and compute expansion.

Patronus combines research-first model evaluation with enterprise product delivery and has raised institutional funding to scale the platform.

Patronus combines research-first model evaluation with enterprise product delivery and has raised institutional funding to scale the platform.

Patronus combines frontier-AI evaluation research with production-grade tooling across hallucination detection, multimodal judging, RAG and agent evaluation, security risks, alignment, custom evaluators, cloud/on-prem deployment and agent-trace debugging, spanning research, ML evaluation, AI reliability and security engineering.

LLM evaluation, hallucination detection, safety, alignment and benchmark design skills transfer exceptionally broadly across frontier labs and enterprise AI.

LLM evaluation, hallucination detection, safety, alignment and benchmark design skills transfer exceptionally broadly across frontier labs and enterprise AI.