Arize Phoenix

Open-source ML observability

Observability Free (OSS)
Visit Official Site →

What It Is

Phoenix from Arize AI is an open-source observability tool for LLM and ML applications. Built on OpenTelemetry standards, it captures traces from any instrumented app and provides a notebook-friendly UI for exploring LLM behavior, debugging issues, and running evaluations.

How It Works

Phoenix uses OpenTelemetry (OTEL) as its ingestion format, meaning it works with any framework that emits OTEL traces (LangChain, LlamaIndex, DSPy, Instructor, and raw API calls via auto-instrumentors). Phoenix runs as a Python package (for notebook use) or as a standalone server (for production). The dashboard provides trace views, retrieval evaluation, embeddings visualization, and LLM-as-judge evals.

Pricing Breakdown

Completely free and open source. Arize AX (their enterprise platform) is paid but Phoenix itself has no paid tier.

Who Uses It

ML researchers, data scientists working in notebooks, and teams preferring open-standard tooling. Popular in research-oriented organizations.

Strengths & Weaknesses

✓ Strengths

  • Open source
  • OpenTelemetry standard
  • Notebook-friendly
  • Strong embeddings visualization

× Weaknesses

  • More technical setup
  • Research-oriented UX
  • Less polished dashboard

Best Use Cases

ML researchEval pipelinesOpen telemetry stacksNotebook workflows

Alternatives

LangSmith
LangChain's observability platform
Langfuse
Open-source LLM observability
← Back to AI Tools Database