AI Governance & Security Vendor Wiki

3 min read

Arize Phoenix

Researched 2026-06-28. Primary category: llm-observability.

One-liner — Open-source AI observability and evaluation library (Phoenix) from Arize AI, with a commercial enterprise platform (Arize AX) alongside.

What it does — Phoenix traces LLM and agent applications, runs evaluations, surfaces failures (hallucinations, retrieval problems, regressions), and helps teams debug and improve output quality. Arize’s heritage is classic ML observability (drift, performance monitoring), and Phoenix extended that into the LLM/agent world. Phoenix is OSS and self-hostable; Arize AX is the managed, enterprise-scale commercial product.

Where it sits in the stack — The llm-observability layer of the model/prompt tier. Dev/ops-time tracing and eval, not an inline guardrail — but its eval/monitoring helps detect data-leak or injection symptoms. Complements ai-runtime-security and ai-red-teaming.

Deployment & architecture — OpenTelemetry/OpenInference-based SDK instrumentation; Phoenix runs locally or self-hosted, Arize AX is SaaS. Integrations include Azure AI Foundry / AI Studio and most major model and framework providers. Reported >2M monthly downloads for Phoenix (company claim).

Positioning & differentiators — Among the strongest OSS-plus-commercial combos in the category, with deeper roots in production ML monitoring than LLM-native startups. Versus langfuse both are OSS-first and overlap on tracing/evals; Arize brings more mature ML-monitoring lineage. Versus braintrust it is observability-and-monitoring-led rather than eval-first; versus datadog it is AI-specialized rather than a general observability suite. Notably, datadog is itself an investor in Arize’s Series C.

Ownership, funding & M&A — Independent, VC-backed. Series C $70M announced 2025-02-20, led by Adams Street Partners (with M12/Microsoft, Datadog, PagerDuty, OMERS, Sinewave, Industry Ventures, and returning Foundation Capital, Battery, TCV, Swift). ~$135M total across 5 rounds. Founded 2020; HQ Berkeley, CA. No M&A. Confidence: high.

CTO / hedge-fund lens — Day-1 for a shop doing RAG or agents that wants both LLM observability and an upgrade path to production ML monitoring. Phoenix OSS lets you keep traces (potential MNPI) self-hosted; Arize AX is the route when you need scale/support. Strong funding and strategic investors lower continuity risk. Not a governance/SR 11-7 system itself — pair with ai-governance-platform.

Competitors / alternatives — langfuse, langsmith, braintrust, helicone, datadog, comet, fiddler-ai, whylabs.

Open questions / to verify

  • Phoenix OSS vs Arize AX feature boundary (what’s gated to commercial).
  • Independent verification of the >2M downloads figure (vendor claim).

Sources

History

  • [2026-06-28] Stub created from seed registry.
  • [2026-06-28] Researched; confirmed independent (Arize AI, Series C Feb 2025 $70M), clarified Phoenix (OSS) vs Arize AX (commercial), filled founding/HQ/funding/deployment. No M&A. ownership_confidence low→high.

Graph View

Backlinks