starter.wikiSearch⌘K

Open-source LLM observability and evaluation

Langfuse

An open-source LLM engineering platform with tracing, evaluations, prompt management, datasets, Cloud plans, and self-hosted deployment.

Visit official site

Editorial verdict

Choose Langfuse when open-source portability or self-hosting materially affects the observability decision, after modeling ingestion units, retention, storage, evaluation models, and platform operations.1234

Best for

  • Teams wanting open-source tracing and evaluation
  • Products comparing Cloud with self-hosted deployment
  • Organizations treating prompt management as part of the evaluation workflow
1234

Not ideal for

  • Teams unwilling to operate self-hosted data infrastructure
  • Buyers assuming open source eliminates storage or evaluation cost
  • Programs that require unverified Enterprise features in the OSS edition
1234
Main trade-off

Open-source deployment choice increases control and portability while adding ingestion, retention, storage, ClickHouse, upgrade, evaluation-model, and edition responsibility.1234

Product boundary

Whether Langfuse Cloud or self-hosting provides the right tracing, evaluation, prompt, retention, data-access, and operating boundary.

Langfuse combines tracing, evaluation, prompt management, datasets, and experiments. Prompt management is a capability within this observability decision, not a separate Task. Cloud, open-source self-hosting, and Enterprise self-hosting have different retention, feature, and operating boundaries.1234

For: AI product teams wanting an open-source observability and evaluation layer with a real Cloud or self-hosted operating decision

  • The production workload, language, model, provider, or deployment boundary changes
  • Current pricing, retention, data use, policy, or regional support changes
  • Measured quality, latency, reliability, or operating cost no longer fits

Why teams consider Langfuse

  • Integrated workflowTracing, scores, datasets, experiments, and prompts share an LLM engineering model.1234
  • Deployment choiceCloud and self-hosted routes have explicit feature and operating boundaries.1234
  • Usage transparencyOfficial documentation defines billable units and retention behavior.1234

Pricing

Langfuse Cloud combines plan fees, included usage, billable units, retention, and add-ons; self-hosted OSS has no vendor usage fee but carries infrastructure and operations.1

Current official route

Cloud Hobby, Core, Pro, Enterprise, or self-hosted

Verified 2026-07-27: Langfuse Cloud publishes plan and usage boundaries based on billable units, retention, users, and add-ons. Self-hosted OSS avoids a vendor usage fee but requires operated data infrastructure.1

Cloud meter
Billable traces, observations, and scores plus plan boundaries1
Retention
Cloud retention differs by plan1
Self-hosted cost
ClickHouse, database, object storage, compute, upgrades, backups, and support1
Pricing checked View official pricing

Langfuse vs alternatives

LangSmith

Choose when
Choose LangSmith when its tracing and evaluation workflow aligns with the development stack and the team accepts its usage-unit, retention, data, and deployment boundaries.
Avoid when
Teams requiring an OSS-first observability core
Compared with Langfuse
Framework-aligned managed evaluation reduces integration work while adding usage-unit, trace-tier, retention, platform, and Enterprise self-hosting dependence.56

Braintrust

Choose when
Choose Braintrust when evaluation-led development is the operating model and its processed-data, scoring, retention, collaboration, and deployment boundaries fit.
Avoid when
Teams primarily seeking an OSS tracing backend
Compared with Langfuse
A focused evaluation workflow improves repeatability while adding dataset, scoring, processed-data, retention, collaboration, and hosted-platform dependence.78

Arize Phoenix

Choose when
Choose Arize Phoenix when an open-source and OpenTelemetry-oriented tracing and evaluation stack justifies owning deployment, storage, access, upgrades, and evaluation-model cost.
Avoid when
Teams wanting a fully managed default
Compared with Langfuse
Open-source tracing and evaluation maximize inspectability and control while transferring data infrastructure, access, retention, upgrade, evaluation-model, and support responsibility.910

Resources and sources

Official product, documentation, pricing, and policy sources

  • Langfuse Cloud pricing
    Open
  • Langfuse self-hosted pricing and editions
    Open
  • Langfuse data retention
    Open
  • Langfuse billable units
    Open
  • LangSmith pricing
    Open
  • LangSmith administration and retention
    Open
  • Braintrust pricing
    Open
  • Braintrust tracing and instrumentation
    Open
  • Arize Phoenix documentation
    Open
  • Arize Phoenix self-hosting configuration
    Open

Related tasks

  1. 1
    Langfuse Cloud pricing

    Langfuse · Accessed Official

  2. 2
    Langfuse self-hosted pricing and editions

    Langfuse · Accessed Official

  3. 3
    Langfuse data retention

    Langfuse · Accessed Official

  4. 4
    Langfuse billable units

    Langfuse · Accessed Official

  5. 5
    LangSmith pricing

    LangChain · Accessed Official

  6. 6
    LangSmith administration and retention

    LangChain · Accessed Official

  7. 7
    Braintrust pricing

    Braintrust · Accessed Official

  8. 8
    Braintrust tracing and instrumentation

    Braintrust · Accessed Official

  9. 9
    Arize Phoenix documentation

    Arize · Accessed Official

  10. 10
    Arize Phoenix self-hosting configuration

    Arize · Accessed Official