Layer A · LLM Observability & Evaluation
Arize (Phoenix and Arize AX)
Two products: Phoenix, a free self-hosted tracing and evaluation project, and Arize AX, the commercial platform with online evaluation, drift monitoring and enterprise support.
Last checked . Visit Arize
At a glance
- Category
- LLM Observability & Evaluation
- Control layer
- A: Build and run
- Five-Layer model
- Usage Transparency, Output Quality
- Status
- Active
- Typical buyers
- Engineering and platform, AI leadership (CAIO)
- Delivery
- Open source, Self-hosted
- Tracing
- Evaluation
- Cost attribution
Buyer fit
Phoenix suits teams that want tracing, evaluation and experiments without a licence and can run the infrastructure themselves. Arize AX suits organisations that need a supported production system of record, especially where predictive ML and generative AI are monitored together. Arize positions the two as complementary: Phoenix for local development and CI, AX for production. Role in the cost-versus-quality decision: traces carry token cost beside evaluation scores, so a cheaper model or prompt can be tested against the quality it must hold.
Market outlook
Established ML observability vendor extending into LLMs. Stable market position with potential for acquisition by larger enterprise software platforms.
What to check
- Phoenix is free to self-host. Check its licence terms, which a competitor describes as source-available rather than OSI open source.
- Arize AX stores data in Arize's own database, so plan the export path before production traces accumulate there.
- Heritage in classic ML observability gives AX mature drift and performance monitoring.
Sources
The map describes vendors from public sources; it does not rate, rank or recommend them, and no listing is paid for. Confirm pricing and capabilities with the vendor. See how the map is built.
By Ed Epstein, who works at IBM. The postings on this site are my own and don't necessarily represent IBM's positions, strategies or opinions. About and disclosure · Methodology