Arize Phoenix.
Open-source LLM observability — the AI/ML observability standard.
What it is.
Arize Phoenix is the open-source LLM observability tool from Arize AI, the company that became the ML observability standard before LLMs took the category. Phoenix runs locally or in cloud, instruments LLM apps via OpenInference and OpenTelemetry, and ships with eval templates, prompt playgrounds, and embedding analysis. Arize AX is the enterprise SaaS layer above it.
Where it fits.
ML platform teams that already use Arize for traditional model monitoring and want one observability surface across classical ML and LLMs. Strong for regulated enterprises with existing Arize procurement. The OpenInference spec makes Phoenix the de facto open-source observability layer outside the LangChain ecosystem.
- OpenInference spec adopted across the industry
- Runs locally — no SaaS lock-in for development
- Native bridge between classical ML and LLM monitoring
- UI denser than developer-focused tools like Helicone
- Enterprise feature set requires Arize AX upgrade
Frequently asked.
What is OpenInference?
Arize's open specification for LLM telemetry, built on OpenTelemetry. Adopted by LlamaIndex, LiteLLM, and others as the common trace format.
Phoenix or Arize AX?
Phoenix is the open-source local tool. Arize AX is the enterprise SaaS with collaboration, alerting, and governance on top.
Does Phoenix work with classical ML monitoring?
Yes through the Arize platform. Phoenix focuses on LLMs; Arize AX covers both surfaces in one tenant.
