Arize AX
by Arize AI
AI observability and evaluation to make AI work reliably in the real world.
Arize AI builds Arize AX, an enterprise AI engineering platform for observability, evaluation, and improvement of LLM applications, AI agents, and traditional ML models in production. It is complemented by Phoenix, the company's open-source LLM tracing and evaluation toolkit.
At a Glance
- Category
- Governance & Security
- Pricing
- Freemium, Subscription, Usage-based, Contact for pricing
- Target Market
- AI Engineers, Data Scientists, ML Engineers, Enterprise Developers, CTOs
- Founded
- 2020
- Headquarters
- Berkeley, California, USA
- Customers
- Hundreds of enterprise customers including Uber, Chime, eBay, Spotify, PepsiCo, Booking.com, and TripAdvisor
Key Features
- ✓Agent and LLM tracing
End-to-end tracing of agent behavior and LLM calls with span, trace, and session-level visibility built on OpenTelemetry/OpenInference.
- ✓Evaluation framework
Online and offline evaluations at scale to detect regressions, hallucinations, and quality issues across spans, traces, and sessions.
- ✓Alyx AI engineering agent
An embedded AI assistant that helps debug issues and recommends improvements to prompts and agent systems.
- ✓Prompt management and optimization
Tools to test, version, and optimize prompts and agent harnesses before production deployment.
- ✓ML production monitoring
Real-time monitoring with drift detection, embedding analysis, and bias tracing for traditional machine learning models.
- ✓Phoenix open-source toolkit
A free, self-hostable open-source observability and evaluation library built on the same OpenTelemetry standards.
Capabilities
Use Cases
- •Production AI agent debugging
Trace and debug autonomous and multi-agent systems to find and fix failures before they impact users.
- •LLM application evaluation
Continuously evaluate chatbots, RAG systems, and copilots for correctness, relevance, and hallucinations.
- •ML model monitoring
Detect model drift, data quality issues, and bias in deployed machine learning models in real time.
Ideal For
Best For
- ✓LLM and AI agent observability and tracing in production
- ✓Evaluating and debugging generative AI applications and multi-agent systems
- ✓Monitoring traditional ML models for drift, bias, and performance degradation
Integrations
Deployment
Market & Ratings
Hundreds of enterprise customers including Uber, Chime, eBay, Spotify, PepsiCo, Booking.com, and TripAdvisor
Market Analysis
Pros
- ✓Strong, deep observability and evaluation for production AI and agents
- ✓Open-source Phoenix option with no vendor lock-in
- ✓Unified coverage of classical ML, LLMs, and agents
- ✓Broad framework and model integrations
Cons
- ✗Engineering-centric interface assumes technical users
- ✗Steep learning curve and documentation can overwhelm beginners
- ✗Enterprise pricing is significant (est. $50k+/year)
- ✗Two-product split between free Phoenix and commercial AX can create friction
Pricing
Phoenix (Open Source)
$0
- ✓Open-source and self-hostable
- ✓Tracing, evaluation, experimentation
- ✓Prompt iteration
- ✓User-managed resources
AX Free
$0
- ✓25k spans/month
- ✓1 GB ingestion
- ✓15-day retention
- ✓Alyx agent and online evals
- ✓Community support
AX Pro
$50/mo
- ✓50k spans/month
- ✓10 GB ingestion
- ✓30-day retention
- ✓Higher rate limits
- ✓Email support
AX Enterprise
Contact for pricing
- ✓Custom spans and ingestion
- ✓Configurable retention
- ✓Dedicated support and uptime SLA
- ✓SOC 2 and HIPAA
- ✓Self-hosting, data residency, multi-region
AX Pro overages are billed at roughly $10/million additional spans and $3/GB additional ingestion. Phoenix is fully free and open source. Third-party sources estimate AX Enterprise contracts start around $50,000/year. Startup pricing program is available.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
Subscribe