Both are evaluation and observability platforms that now belong to large observability vendors, Arize to Dynatrace and Galileo to Cisco's Splunk. Arize is the one you can start for free, self-hosted or hosted; Galileo is the one with guardrails that act on tool inputs and outputs during a run. The observability stack you already pay for often settles the rest.[3][6]
Arize and Galileo differ most in what you can start with for free and in how far their run-time checks reach. Arize offers Phoenix, free to self-host, and a free hosted AX tier; Galileo, sold as Splunk Agent Observability since 7 August 2026, lists only paid annual pricing. Galileo's Agent Control can block harmful model and tool inputs and outputs during a run, while Arize's guards run in your own code around model calls. Both changed owners in 2026: Cisco completed its Galileo deal in May and Dynatrace completed its Arize deal on 1 October.
| Decision factor | Arize | Galileo |
|---|---|---|
| Ownership | Dynatrace, acquisition completed 1 October 2026; Phoenix and AX continue | Cisco, acquisition completed May 2026; sold as Splunk Agent Observability since 7 August 2026 |
| Products | Phoenix (self-hosted) and Arize AX (hosted, or self-hosted on Enterprise) | One platform, SaaS in Splunk Observability Cloud or on-premises |
| Licence | Phoenix: Elastic License 2.0 (source-available); AX: commercial | Proprietary platform; SDKs and Agent Control are Apache-2.0 |
| Run-time guardrails | Partial: AX guards run in your code via Guardrails AI | Yes: Agent Control checks model and tool inputs and outputs during a run |
| Low-cost judges | Decision Judge for high-volume structured evals | Luna-2 small models; SaaS edition limited to five preset Luna evaluators |
| Self-hosting | Phoenix free; AX on Kubernetes, including air-gapped, on Enterprise | On-premises, including air-gapped (commercial) |
| Free tier | AX Free: 25k spans a month, 15-day retention; Phoenix free to self-host | None listed for Agent Observability |
| Paid pricing | AX Pro: $50 a month, 50k spans | From $100 a month per 1.2 million spans a year, billed annually |
| Capability | Arize | Galileo | Prefactor |
|---|---|---|---|
| Observe | |||
| LLM call tracing | ✓ | ✓ | ✓ |
| Multi-step agent tracing (tool calls, nested spans) | ✓ | ✓ | ✓ |
| Cost tracking | ✓ | ✓ | ✓ |
| OpenTelemetry support | ✓ | ✓ | ✓ |
| Evaluate | |||
| Offline evals (datasets, experiments) | ✓ | ✓ | ✓ |
| Online evals on production traffic | ✓ | ✓ | ✓ |
| LLM-as-a-judge scoring | ✓ | ✓ | ✓ |
| Code-based evaluators | ✓ | ✓ | ✓ |
| Human review and annotation | ✓ | ✓ | ✓ |
| Prompt versioning and playground | ✓ | Partial | — |
| Act on findings | |||
| Runtime controls on live agent actions | Partial | ✓ | ✓ |
| Gateway or proxy in the request path | Partial | — | — |
| Licence and deployment | |||
| Open source | Partial | Partial | Partial |
| Self-hosting | ✓ | ✓ | Partial |
Cisco completed its acquisition of Galileo in May 2026, and since 7 August 2026 the product has been sold as Splunk Agent Observability inside Splunk Observability Cloud. Customers onboarded before that date still use Galileo's own documentation and console; later customers use Splunk's. The old Galileo pricing page now redirects to Splunk.
Dynatrace completed its acquisition of Arize on 1 October 2026. Its announcement says both Phoenix and AX will continue and that Arize capabilities will be integrated into Dynatrace over time, so packaging may change. Galileo is further along that path: it already carries the Splunk name.
In practice this pair often comes down to which parent you already use. A team that sends infrastructure telemetry to Splunk gets agent evaluation in the same place with Galileo. A Dynatrace customer can expect Arize to move closer to its existing tooling. If you use neither, compare the products on their own terms and allow for further change in both.
Arize gives you two free routes. Phoenix is free to self-host with pip, Docker or Kubernetes and covers tracing, datasets, experiments, LLM and code evaluators and prompt versioning. AX Free is a hosted tier with 25k spans a month and 15-day retention. The catch is that continuous evaluation of production traffic, monitors and alerting are AX features; Phoenix's own documentation points to AX for them. Phoenix's Elastic License 2.0 also forbids offering it to others as a hosted service.
Galileo has no free tier listed for Agent Observability. Splunk offers a separate free edition of Observability Cloud, but its coverage of this product is not stated. Purchasing is annual, and on-premises deployment, including air-gapped, is a commercial option rather than an open-source release.
Instrumentation carries across. Both accept OpenTelemetry traces, and Arize maintains OpenInference, which Galileo also reads through its Python instrumentors. Both list integrations for the OpenAI Agents SDK, LangGraph, CrewAI, Google ADK, Mastra and the Vercel AI SDK, which reduces the work of trialling one and moving to the other. Galileo adds Strands, Microsoft Agent Framework and A2A multi-agent tracing; AX adds an agent trajectory graph that maps which agents call which.
This is the clearest product difference. Galileo's Agent Control evaluates model and tool inputs and outputs while an agent runs and can block prompt injection, personal data leaks and harmful content using rules set centrally, without changes to agent code. It replaced the deprecated Protect product in June 2026, and its core was released under Apache-2.0 in March 2026. Blocking content is documented; approval workflows for actions are not.
Arize AX documents guards built on the Guardrails AI framework. They run inside your own code around model calls and can return a default response, re-ask or raise an exception, and guard failures can be traced and alerted on. The platform itself observes rather than intercepts traffic, and there is no documented policy layer for agent tool actions.
On evaluators, Galileo offers Luna-2 small models as lower-cost judges, but the Splunk-hosted edition supports only five preset Luna evaluators and no custom code evaluators; those remain on-premises only. Arize offers LLM judges, agent-as-a-judge on every plan and code evaluators in both products.
As of 20 October 2026, Arize AX Free costs nothing for 25k spans a month with unlimited users, and AX Pro is $50 a month for 50k spans, 10 GB of ingestion and 30-day retention. AX Enterprise is custom and is the only way to self-host AX. Splunk lists Agent Observability from $100 a month per 1.2 million spans a year, billed annually, with an allowance of Luna tokens and fine-tuning runs.
The units differ: Arize prices a monthly span allowance, Splunk an annual one. Model your own traffic against Arize's pricing and Splunk's pricing, and expect both to be revisited as the products fold into their new parents.
We sell the layer this section describes. Read it with that in mind.
Stated plainly, since Prefactor publishes this page: Prefactor is an agent evaluation platform that scores each step of a run and acts on what it finds through runtime policies on tool calls and data access, approval holds for risky actions, and a kill switch for a run, an agent or a team. Galileo's Agent Control is the nearer of these two to that, since it blocks content during a run; Arize's guards sit in your own code around model calls. Neither documents approval holds. Prefactor has no prompt playground or prompt registry and its platform is hosted, so if free self-hosting is the requirement, Phoenix is the better fit. See the full comparison of 15 platforms.
Reviewed against public product and documentation pages on October 20, 2026. If a vendor has changed a feature, product name, or positioning since then, send a correction and we will update it. Numbered source links in the page body point to the ordered sources below.
Prefactor scores each step of every agent run and can hold or stop an action when a score says it should not happen. Start on the free tier: 5,000 spans a month, no credit card required.
Prefactor helps teams observe, evaluate, and improve their AI agents in production, across every framework and provider.