Langfuse vs. Arize AX / Arize Phoenix
This guide outlines the key differences between Langfuse, Arize AX, and Arize Phoenix to help engineering teams choose the right LLM observability platform. Product facts below were checked against public Arize sources on August 13, 2026.
In August 2026, Dynatrace announced a definitive agreement to acquire Arize. As of publication, Arize AX and Phoenix remain available. This page compares the current products.
TL;DR:
- Choose Langfuse if you want MIT-licensed tracing and evaluation, self-hosted or on Cloud, with usage-based Cloud pricing and no vendor lock-in. It is the open-source AI engineering platform with the most public adoption.
- Choose Phoenix if you want an Elastic License 2.0, local-first platform for tracing, evaluations, datasets, experiments, and prompt management. It can run as one container (SQLite) or multi-instance deployments (PostgreSQL).
- Choose Arize AX if you want Arize's managed production platform, listed PCI DSS 4.0 status, an Enterprise self-hosted option, and Enterprise Data Fabric access, which is currently waitlisted.
Already on Phoenix? See Migrate from Arize Phoenix to Langfuse. Already on AX? See Migrate from Arize AX to Langfuse.
Talk to us if you want help with the move.
Open Source & Distribution
Langfuse is MIT-licensed across tracing, evaluations, prompt management, and experiments. OSS self-hosting runs the same core product and ClickHouse architecture as Cloud, so you can move between Cloud and self-host, or off Langfuse, without rewriting the application layer. Enterprise governance modules (audit logs, project-level RBAC, SCIM, protected labels, retention policies) need an Enterprise license when self-hosted.
Arize ships two products. Phoenix is source-available under the Elastic License 2.0 (not an OSI-approved open-source license) and self-hosts on SQLite or PostgreSQL. Arize AX is proprietary SaaS (self-hosting is an Enterprise option). Phoenix and AX do not share a storage engine: Phoenix uses Postgres/SQLite; AX Cloud uses adb.
Dynatrace's acquisition announcement is an enterprise AI observability deal. It refers to Arize as "OSS-native" and highlights its open-source developer community, but it does not name Phoenix, and it does not state a license or a commitment that a standalone open-source product continues. For current Phoenix licensing and availability, see the Phoenix license and self-hosting documentation.
| Feature | Langfuse | Arize AX |
|---|---|---|
| Model | Open Source (MIT License) | AX: proprietary SaaS. Phoenix: Elastic License 2.0 (source-available) |
| GitHub Stars |  |  |
| PyPI Downloads |  |  |
| npm Downloads |  |  |
| Docker Pulls |  |  |
| Self-Hosting | Free OSS: same codebase and ClickHouse architecture as Cloud. Enterprise governance needs an Enterprise license. | Phoenix: self-host on SQLite/Postgres. AX: self-host on Enterprise. |
Scalability & Performance
Both tools are built for scale, but they use different architectural approaches. Langfuse is part of ClickHouse and leverages the speed of ClickHouse architecture, while Arize AX uses a proprietary database.
| Feature | Langfuse | Arize AX |
|---|---|---|
| Backend | ClickHouse (acquired Langfuse): Optimized for high-throughput OLAP. | adb (Arize Database): Proprietary engine for agentic telemetry. |
Integrations
Both platforms ingest OpenTelemetry. Langfuse maps OpenInference attributes natively, so existing OpenInference instrumentors can keep running and export to Langfuse.
| Feature | Langfuse | Arize AX / Phoenix |
|---|---|---|
| Standard | OpenTelemetry Native: Built on OTel standards. | OpenTelemetry compliant: Built on OTel standards. |
| Integrations | 120+ integrations plus OpenInference ingest via OTLP. | OpenInference auto-instrumentation across popular frameworks. |
Pricing
Langfuse Cloud meters one unit (a trace, observation, or score) with no seat fees. Arize AX meters spans and ingested GB, with retention included per plan. Phoenix self-hosted has no platform usage fee; you run the infrastructure. Figures below are from Arize pricing and Langfuse pricing as of August 2026.
| Feature | Langfuse | Arize AX |
|---|---|---|
| Model | Usage-Based: Billable unit = trace, observation, or score. | Hybrid: spans + ingestion volume (GB). |
| Free Tier | 50k units/mo, 30-day retention, 2 users. | 25k spans/mo, 1 GB, 15-day retention. |
| Paid entry | Core $29/mo (100k units, 90-day retention). | Pro $50/mo (50k spans, 10 GB, 30-day retention). |
| Retention | 3 years on Pro and Enterprise. | 15 days Free, 30 days Pro, custom on Enterprise. |
| Self-host | Core platform on free OSS; enterprise governance needs an Enterprise license. | Phoenix: free to self-host. AX: Enterprise. |
| Plans | Hobby (free), Core ($29/mo), Pro ($199/mo), optional Teams add-on ($300/mo), Enterprise ($2,499/mo). | Free, Pro ($50/mo), Enterprise. Phoenix is separate and self-hosted. |
Open Platform & Extensibility
Langfuse is designed as a core infrastructure component, allowing teams to build custom internal tools on top of its API.
| Feature | Langfuse | Arize AX |
|---|---|---|
| API Access | API first for all data (traces, evals, prompts) and platform features. | API available, exposes only core resources. |
| Customizability | Build custom workflows, evaluations, and dashboards using the SDK/API. | Custom evaluations and pipelines via SDK. |
| Data Access | Query via API on all plans. Scheduled blob exports: Teams add-on, Enterprise, or self-hosted. | Export via UI, SDK, API, CLI. |
Enterprise Security
Both platforms serve large enterprises. Arize AX lists PCI DSS 4.0 among its certifications. Langfuse lists SOC 2 Type II and ISO 27001, with a HIPAA-ready region on Pro or higher, and supports masking to redact sensitive fields before ingest.
| Feature | Langfuse | Arize AX |
|---|---|---|
| Certifications | SOC 2 Type II, ISO 27001, GDPR, HIPAA-ready region (Pro+). | SOC 2 Type II, HIPAA, PCI DSS 4.0, CSA Star Level 1. |
| Regions | Cloud EU, US, Japan; HIPAA on Pro+. | Cloud US, EU, or CA (pricing). |
| Adoption | Trusted by 21 of Fortune 50 & 129 of Fortune 500. | Strong enterprise adoption, including teams that already run Arize for classical ML. |
| Governance | Org RBAC, Enterprise SSO and project RBAC: Teams add-on or Enterprise. Audit logs: Enterprise. Self-host OSS includes org RBAC and SSO; project RBAC and audit logs need Enterprise. | Organization/space RBAC with plan-specific scope; SSO, fine-grained space-level RBAC, audit logs on Enterprise (pricing). |
Feature Highlights
Langfuse:
- Core Observability: Best-in-class tracing with accurate token and cost tracking for 100+ models.
- Prompt Management: Collaborative playground with versioning, caching, fallbacks, and protected labels.
- Collaboration: Annotation queues, comments with @mentions, and audit logs.
- Evaluations: Flexible "LLM-as-a-Judge" evaluators that can be run in-UI or via SDK pipelines.
Arize AX:
- Agentic Visualization: Specialized views for multi-agent conversation flows.
- Data Fabric: Sync with enterprise data lakes (Enterprise feature, currently waitlisted).
- Evaluation: Session-level evaluation and retrieval diagnosis (RAG). AX also includes Signal and Alyx, with availability depending on plan.
Already instrumented with OpenInference? Keep it and follow the Arize AX migration or the Phoenix migration. For a migration plan on Cloud or self-hosted, talk to us.
This comparison is out of date? Please raise a pull request with up-to-date information.
Last edited
