Patronus AI
Patronus AI is an automated AI evaluation and security platform offering Digital World Models for agent simulation, hallucination detection via the Lynx model, and benchmarking tools. It serves frontier AI labs and enterprises deploying large language model applications, founded in 2023 by former Meta AI researchers.
- Company typePrivate
- Founded2023
- HeadquartersSan Francisco, United States
- Headcount51–100
- GTM typeB2B
- OfferingSoftware
What Patronus AI does
Patronus AI is an automated AI evaluation and security platform that helps enterprises and frontier AI labs deploy large language models and AI agents safely. Founded in 2023 by former Meta AI researchers Anand Kannappan (CEO) and Rebecca Qian (CTO), the company is headquartered in San Francisco and operates with 11-50 employees. The company's stated mission is to develop simulation research and infrastructure to accelerate progress toward human-aligned AGI. Its customer base spans frontier AI labs (including OpenAI), technology companies (HP, Databricks, Cohere), education/publishing (Pearson), and e-commerce (Etsy), with the company claiming to work with the majority of the world's leading frontier AI labs.
The platform's core technology centers on Digital World Models — large-scale simulation environments that predict and simulate agent actions in digital workflows, positioned as foundational infrastructure for self-adaptive worlds supporting continual learning. Complementary proprietary products include Lynx (a 70B-parameter SOTA hallucination detection model that outperforms GPT-4, with an 8B Lynx v1.1 variant for RAG tasks), Glider (a 3B open-source SLM judge for explainable evaluations), Percival for AI agent evaluation and benchmarking, FinanceBench (10,000 Q&A pairs on financial documents), BLUR (573 tip-of-the-tongue Q&A pairs), Generative Simulators with adaptive environment generation, the Open Recursive Self-Improvement (ORSI) training method, and the industry's first multimodal LLM-as-a-judge for image-to-text evaluation. The platform also offers Patronus Experiments, Logs, Traces, Comparisons, and Datasets for the full evaluation workflow.
Patronus AI operates a hybrid GTM model combining enterprise field sales (free 45-minute evaluation consultations targeting CTOs and AI Leads, custom Enterprise pricing with on-premise and dedicated VPC deployment) with self-serve product-led growth (free Individual tier, Base tier at $25/month, and a developer API launched in 2024 with $10 free credits). Revenue streams include tiered SaaS subscriptions, usage-based API pricing ($10-$20 per 1K calls), and custom AI services (eval model fine-tuning, eval dataset generation). The platform holds SOC 2 Type II, TISAX, and HIPAA certifications and is distributed through partnerships with AWS, Databricks, MongoDB, and Hugging Face. Total capital raised is $70M across three rounds led by Lightspeed Venture Partners (seed), Notable Capital (Series A), and Greenfield Partners (Series B), with strategic participation from Datadog, Samsung, and Factorial Capital.
Patronus AI firmographics
Firmographics- Name
- Patronus AI
- Legal name
- Patronus AI, Inc.
- Website
- https://patronus.ai
- Company type
- Private
- Founded year
- 2023
- Operating status
- Operating
- Headcount range
- 51–100 employees
- Short description
- Patronus AI is an automated AI evaluation and security platform offering Digital World Models for agent simulation, hallucination detection via the Lynx model, and benchmarking tools. It serves frontier AI labs and enterprises deploying large language model applications, founded in 2023 by former Meta AI researchers.
- Ownership category
- akta.pro rank
Patronus AI industry classification
Industry- Product category
- AI Evaluation and Safety Platform
- NAICS
- Software Publishers (5132), Computer Systems Design and Related Services (54151)
- SIC
- Services-Prepackaged Software (7372)
- akta.pro primary industry
- Responsible AI, Security & Privacy Platforms (Safety, Guardrails, PII) (HDAEANAG)
- akta.pro secondary industries
- AI Observability, Monitoring & Evaluation Platforms (Drift, Quality, Safety) (HDAEANAF), Responsible AI Policy, Controls & Workflow Management (HDAAAMAG), Model Hosting, Serving & Inference Platforms (HDAAACAB)
Keywords
Where Patronus AI is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Patronus AI business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Infrastructure, Marketing or Sales, Operations
Revenue model
- SaaS Platform Subscription: Tiered SaaS platform subscription model with Individual (free), Base ($25/month), and Enterprise (custom pricing) tiers. Platform includes evaluation runs, logs, traces, comparisons, datasets, and API access.
- API Usage-Based: Usage-based API pricing for evaluator calls. $10 per 1k small evaluator API calls, $20 per 1k large evaluator API calls, $10 per 1k eval explanations. Developers get $10 in free credits to start.
- Enterprise Custom Solutions: Custom AI services including eval model fine-tuning and eval dataset generation offered as part of Enterprise tier. On-premise and dedicated VPC deployments with custom data retention and SSO capabilities.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Monthly | Individual — Free tier for individual developers |
| Subscription | Monthly | Base — Mid-tier subscription with expanded access |
| Subscription | Multi-year contract | Enterprise — Unlimited custom deployment |
| Usage-based | Pay-as-you-go | Developer API — Pay-as-you-go API usage |
Go-to-market motion3 records
Distribution channels5 records
Marketing channels8 records
Patronus AI product offering
Product offeringCore offering
Patronus AI develops an automated evaluation and security platform for large language models (LLMs) and AI agents. Its core products include the Core Platform (Patronus Experiments, Logs, Traces, Comparisons, Datasets), Percival for AI agent evaluation, RL Environments, Digital World Models, and Generative Simulators for training and evaluating AI agents in adaptive digital workflows. The company also builds proprietary evaluator models such as Lynx (hallucination detection), Glider (small language model judge), and the Multimodal LLM-as-a-Judge, alongside benchmarks like FinanceBench and BLUR.
Product overview
Patronus AI is a unified platform with multiple integrated products for AI evaluation and agent training. The core offering includes the Core Platform for automated LLM evaluation and security testing, supplemented by specialized models like Lynx (hallucination detection) and GLIDER (explainable evaluation). The company has expanded into simulation infrastructure with Digital World Models and Generative Simulators for AI agent training. Additional components include RL Environments for reinforcement learning, the Patronus API for programmatic access, and benchmarking tools like FinanceBench and BLUR. The product architecture supports enterprise evaluation needs through tiered pricing (Developer, Enterprise) with features like on-premise hosting, SSO, and custom data retention options.
Differentiator
Problem solved
Functional benefit
Products and services
- Patronus AI Core Platform The flagship SaaS platform that provides automated model scoring, adversarial test generation, benchmarking, evaluation runs, logs, traces, comparisons, and dataset management for enterprises deploying LLMs. Targeted at enterprise AI teams and AI labs.
- Percival AI agent evaluation and benchmarking product within the Patronus platform, designed for evaluating and training AI agents.
- RL Environments Reinforcement learning environments for training and evaluating AI agents across domains including software development, customer service, and finance.
- Digital World Models Large-scale simulation environments that predict and simulate agent actions in digital workflows, used to scale creation of high-alpha simulations for frontier model training and continual learning.
- Generative Simulators Adaptive simulation environments that dynamically generate tasks, rules, and evaluations to evaluate AI agent behavior over time, replacing static benchmarks for agent training.
- Lynx State-of-the-art hallucination detection model (70B parameters) that achieves the highest accuracy at detecting AI hallucinations, outperforming GPT-4 on hallucination tasks. Available off-the-shelf and as a custom evaluator; also open-sourced.
- Glider Open-source 3B small language model (SLM) judge for fast and explainable AI evaluations. Generates structured reasoning chains with confidence scores and contextual highlights for cost-effective guardrails.
- Multimodal LLM-as-a-Judge Industry's first multimodal judge model designed to evaluate and optimize AI systems that convert image inputs into text outputs (vision-language models).
- Patronus API Self-serve API for AI evaluation, hallucination detection, and guardrails with usage-based pricing. Enables developers to integrate evaluation capabilities programmatically with a single line of code.
Quantifiable outcome
- Revenue grew more than 15x over the past year (as of June 2026 Series B)
- +5 more outcomes
Companies that use Patronus AI
Customer profileNamed customers7 records
Segments6 records
Ideal customer profiles4 records
Patronus AI technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability8 records
Feature10 records
Patronus AI partnerships and signals
Strategic signalPartnerships
Four partnerships are on record, tiered secondary and primary.
- Hugging FacesecondaryPatronus AI's open-source tools including Lynx and Glider are hosted and distributed through Hugging Face. Glider is hosted at Hugging Face as an open-source SLM judge model. The OpenEnv project transitioned to community governance with a coordination committee including Meta-PyTorch, Nvidia, Microsoft, Hugging Face, and others.
- Amazon Web Services (AWS)primaryAWS provides cloud infrastructure for Patronus AI's platform operations. The company uses AWS for infrastructure management and conducts periodic architecture and security reviews with AWS. Featured in AWS case study titled 'How Patronus AI helps enterprises boost their confidence in generative AI'. Patronus AI is also vetted by AWS as part of their partner ecosystem.
- DatabricksprimaryStrategic partnership and integration within the Databricks ecosystem. Patronus AI has a dedicated case study page at patronus.ai/databricks. Databricks is listed as a partner for joint go-to-market and customer deployment. Patronus is integrated into Databricks' partner ecosystem.
- MongoDBsecondaryTechnology integration partner. MongoDB is listed as a partner alongside AWS and Databricks on the Patronus AI consultation page.
Scale indicators8 records
Recent moves6 records
Expansion highlights6 records
Patronus AI competitors and assessment
Company assessmentMarket position
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
Patronus AI social profiles
Digital presencePatronus AI compliance and trust
Trust signalCompliance3 records
Patronus AI financial estimates
Financial estimateRevenue estimate
Valuation estimate
Patronus AI leadership team
Management profileNumber of profiles
Profiles2 records
Patronus AI funding detail
Funding detailFunding overview
Funding rounds3 records
Investors8 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Patronus AI M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Patronus AI
What does Patronus AI do?
Patronus AI develops an automated evaluation and security platform for large language models (LLMs) and AI agents. Its core products include the Core Platform (Patronus Experiments, Logs, Traces, Comparisons, Datasets), Percival for AI agent evaluation, RL Environments, Digital World Models, and Generative Simulators for training and evaluating AI agents in adaptive digital workflows. The company also builds proprietary evaluator models such as Lynx (hallucination detection), Glider (small language model judge), and the Multimodal LLM-as-a-Judge, alongside benchmarks like FinanceBench and BLUR.
Is Patronus AI a public or private company?
Patronus AI is a private company. It is classified as venture growth investor backed and is currently operating.
When was Patronus AI founded?
Patronus AI was founded in 2023. It employs 51 to 100 people.
Where is Patronus AI based?
Patronus AI is headquartered in San Francisco, United States, in the North America region.
How does Patronus AI make money?
Three revenue lines are on record. SaaS Platform Subscription is the primary driver. The others are API Usage-Based and enterprise Custom Solutions.
Does Patronus AI have an API?
Yes. Self-serve API for AI evaluation and guardrails, including hallucination detection. Provides small evaluator API calls at $10/1k calls and large evaluator API calls at $20/1k calls. Also offers eval explanations at $10/1k. Enterprise tier includes higher rate limits and volume discounts. Developer documentation is at docs.patronus.ai/docs.
What industry is Patronus AI in?
Patronus AI's product category is AI Evaluation and Safety Platform. Its primary akta.pro industry code is HDAEANAG, Responsible AI, Security & Privacy Platforms (Safety, Guardrails, PII), with a secondary code of HDAEANAF, AI Observability, Monitoring & Evaluation Platforms (Drift, Quality, Safety). Its NAICS code is 5132 and its SIC code is 7372.