Pick MLflow Tracing if
teams already on MLflow who will run Tracing locally at $0 (uvx mlflow demo or their own server) — not Weave first unless they want W&B Pro $60.
Read the MLflow Tracing pageMLflow Tracing vs Phoenix
Our take
Pick MLflow Tracing if you're on a team already on MLflow who will run Tracing locally at $0 (uvx mlflow demo or their own server) — not Weave first unless they want W&B Pro $60. Pick Phoenix if you're on a agent team who will self-host Phoenix for $0 (uvx arize-phoenix serve) and only open Arize AX when they want a managed cloud.
Best for
MLflow Tracing LLM traces next to the MLflow you already run, at $0
Phoenix See why an agent failed and prove a fix
Free plan
MLflow Tracing $0 · local / your server. You bring machines
Phoenix $0 · local / Docker / Kubernetes. You bring compute. Homepage: uvx arize-phoenix serve
Lowest paid
MLflow Tracing Not published by the vendor
Phoenix AX Pro $50/mo
Runs where
MLflow Tracing Hosted
Phoenix Hosted
Watch out
MLflow Tracing $0 is the OSS server, not your warehouse
Phoenix ELv2, not Apache
Last checked
MLflow Tracing 2026-10-27 · mlflow.org/docs/latest/genai
Phoenix 2026-10-27 · arize.com/pricing
Which one, then?
Pick Phoenix if
agent teams who will self-host Phoenix for $0 (uvx arize-phoenix serve) and only open Arize AX when they want a managed cloud.
Read the Phoenix page