Galileo vs Arize Phoenix in 2026 - Enterprise Eval Intelligence vs Open-Source OTel

Galileo and Arize Phoenix both do eval and observability, but they are aimed at completely different buyers, which is why comparing them is genuinely useful. Galileo is the enterprise eval-intelligence platform - research-grade metrics, proprietary Luna models, real-time guardrails, and a sales rep on the other end of the phone. Phoenix is the open-source, OpenTelemetry-native option you spin up on your laptop for free. One is a platform you buy through a procurement process; the other is a pip install. The right pick is mostly about whether you are an enterprise with a hallucination problem or a team that wants to start tracing today. I have added Langfuse as a third open-source reference.

The short version

Tool	Best for	Self-host	Pricing model	Starting price
Galileo	Enterprise eval intelligence + guardrails	Enterprise only	Free tier, then sales-led	Free / $100/mo
Arize Phoenix	Fast OSS tracing and RAG eval	Full, free	Free OSS	Free (OSS)
Langfuse	Cheap self-host, framework-agnostic	Full, free	Transparent, usage-based	Free / $29/mo

Galileo: research-grade evals and real-time guardrails

Before anything else, the disambiguation. There are two unrelated companies using the Galileo name. The one here is the LLM eval platform at galileo.ai, founded in 2021 by Vikram Chatterji, Atindriyo Sanyal and Yash Sheth. A different product, Galileo AI at usegalileo.ai, was a text-to-UI design tool Google acquired. Verify the domain on every source before you trust a number.

With that cleared up, Galileo is the best-funded platform in this space at roughly $68M raised, including a $45M Series B in October 2024. Its technical bet is proprietary Luna and Luna-2 eval models - small models fine-tuned for tasks like hallucination and prompt-injection detection - cheap and fast enough to run on every request, which is what turns offline evals into real-time production guardrails. Galileo claims Luna is up to 11x faster and 97% cheaper than a GPT-3.5-based judge, but those are vendor benchmarks - treat them as a claim until you test on your own traffic. The free tier is genuinely generous: 5,000 traces, unlimited users, unlimited custom evals.

The reservations are commercial. Above the $100/mo Pro tier (50,000 traces, billed yearly) everything is contact-sales, and self-host is Enterprise-only with no open-source version. It supports OpenTelemetry across CrewAI, LangGraph and the OpenAI Agents SDK. The recurring, credible complaint is that you cannot size ROI without talking to a rep. This is a platform you buy through a sales process, not a card.

Arize Phoenix: fastest to try, best RAG eval, read the license

Arize Phoenix is the OSS pick that starts fastest - a working trace UI on your laptop in under a minute. It is genuinely OpenTelemetry-native, built on OTel and Arize’s own OpenInference conventions, so it is framework-agnostic. Its 50+ pre-built eval metrics include the best RAG evaluation in the category - serious retrieval and answer scoring without writing your own judge prompts. Phoenix OSS is free with no usage caps and no feature gates on the actual features, and Arize is well funded (a $70M Series C in February 2025).

The gotcha is the license. Arize markets Phoenix as “fully open source, no feature gates,” but the main server repo is Elastic License 2.0 - source-available, not OSI-approved open source. ELv2 forbids offering Phoenix as a hosted service to third parties. Only the client and eval subpackages are Apache-2.0. For internal use it behaves like open source and the features are not gated; if your plan is to resell it as a service, the license rules it out. There are also reports of ingest lag before traces appear. And Arize AX, the separate managed cloud, has pricing you cannot read publicly - budget a sales call there.

Where Langfuse fits

Both Galileo and Phoenix have a catch - Galileo hides real pricing behind sales, Phoenix ships a source-available server. Langfuse is the option with neither problem: MIT-licensed, self-hosts free with only three features enterprise-gated, and transparent usage-based cloud pricing that runs about $101/mo at 1M events. It does not have Galileo’s Luna guardrails or Phoenix’s category-best RAG eval, and the v3 self-host is a four-service stack. But if what you want is a cleanly-licensed, transparently-priced open-source platform, it is the safer default than either.

Galileo vs Arize Phoenix: which should you pick?

• You are an enterprise that got burned by production hallucinations - Galileo, for the Luna-plus-guardrails loop, and confirm you have the right company before you trust any price.
• You want to start tracing in the next 60 seconds, for free - Arize Phoenix, local UI in under a minute.
• RAG evaluation is the priority - Phoenix has the best in the category.
• You need real-time production guardrails on every request - Galileo’s Luna models are built for exactly that; Phoenix does eval but not the same live-guardrail loop.
• You need transparent pricing you can commit to on a card, or free open-source self-host - avoid Galileo’s sales motion; Phoenix or Langfuse.
• You need strict OSI-approved open source - Langfuse, not Phoenix’s ELv2 server.

The honest summary: these two barely compete for the same buyer. Galileo is an enterprise purchase with genuine research behind it and a sales process to match. Phoenix is a free, fast, OTel-native tool with a license caveat that only matters if you resell it. If you have a procurement team and a production hallucination problem, Galileo. If you want to self-serve and start today, Phoenix - or Langfuse for a cleaner license.

Every price and date here was read from each vendor’s own materials and verified on 26 July 2026. Luna benchmark figures are Galileo’s own claims. This category ships breaking changes monthly, so we re-verify every 30 days.