Langfuse is the best LLMOps platform: it combines tracing, prompt management and evaluation, is MIT licensed apart from its ee folders, and offers a free Hobby plan with 50k units per month, then $29 per month for Core. LangSmith is #2 and Braintrust is #3. Features and prices come from each vendor’s own pages.
Full visibility into an LLM application, from individual traces to production-wide metrics, with evaluation, prompt engineering and deployment in every platform option
LangChain and LangGraph teams that also want a managed deployment option
Tracing built on OpenTelemetry for any LLM provider and agent framework, evaluation with 50+ built-in metrics and LLM judges, prompt registry and an AI Gateway
Teams that want one open-source tool for classic ML tracking and LLM tracing under a permissive license
A summary of AI metrics from response time and quality to tokens, APM golden signals and infrastructure
Teams already on New Relic who want AI metrics next to application performance data
Pick by what you want
Best overall
Langfuse
Best for LangChain teams
LangSmith
Best for eval-first teams
Braintrust
Best for existing Datadog shops
Datadog LLM Observability
Best cheapest paid cloud plan
Comet Opik
Best open source for classic ML too
MLflow
The ranking
1
Langfuse
Answrs VerdictOpen source, self-hostable and priced in public from $0.
Headline feature: Tracing built on OpenTelemetry, prompt management, evaluation and datasets, with a public API; the page calls it fully open sourcePrice: Hobby is free with 50k units per month; Core is $29 per month and Pro is $199 per month, each with 100k units, then $8 per 100k unitsRetention: Data access is 30 days on Hobby, 90 days on Core and 3 years on ProOpen source: The repository is MIT licensed, except for the ee foldersOwnership: The repository says that since January 2026 Langfuse has been part of ClickHouse
Langfuse covers tracing built on OpenTelemetry, prompt management, evaluation with LLM-as-a-judge and annotation queues, datasets and a playground. Hobby is free with 50k units per month, Core is $29 per month and Pro is $199 per month, each paid plan with 100k units and then $8 per 100k. Data access runs 30 days on Hobby, 90 days on Core and 3 years on Pro. The repository is MIT licensed except the ee folders and says Langfuse has been part of ClickHouse since January 2026.
Go ifTeams that want tracing, prompt versioning and evals in one open-source tool they can self-host
Skip ifYou want a vendor that does not meter usage in units
Why #1 over #2: It publishes a $29 Core plan and an MIT repository, where LangSmith’s Plus plan is $39 per seat.
2
LangSmith
Answrs VerdictThe most complete managed platform if your stack already runs on LangChain.
Headline feature: Full visibility into an LLM application, from individual traces to production-wide metrics, with evaluation, prompt engineering and deployment in every platform optionPrice: Developer is $0 per seat per month and Plus is $39 per seat per month, then pay as you go; Enterprise is customIncluded traces: 5k base traces per month on Developer and 10k on PlusRetention: Base traces are kept for 14 days and extended traces for 180 daysCompatibility: LangSmith works with many frameworks and providers, including OpenAI, Anthropic, CrewAI, Vercel AI SDK and Pydantic AI
LangSmith gives full visibility into an LLM application from individual traces to production-wide metrics, with evaluation, prompt engineering and deployment in every platform option. Developer is $0 per seat per month and Plus is $39 per seat per month, then pay as you go, with 5k and 10k base traces per month. Base traces are kept 14 days and extended traces 180 days. It works with OpenAI, Anthropic, CrewAI, Vercel AI SDK and Pydantic AI.
Go ifLangChain and LangGraph teams that also want a managed deployment option
Skip ifYou want pricing that does not scale per seat
Why #2 over #3: It bundles deployment, but it charges per seat and keeps base traces 14 days, against 30 to 90 days of data access on Langfuse’s lower plans.
3
Braintrust
Answrs VerdictEval-first, with unlimited users even on the free plan.
Headline feature: Evals to compare prompts and models, logging and tracing, datasets, and Loop, an AI agent that analyzes data and builds evalsPrice: Starter is $0 per month; Pro is $249 per month; Enterprise is customStarter limits: 1 GB of processed data, 10k scores and 14-day data retention, with unlimited users, projects, datasets and experimentsPro limits: 5 GB of processed data, 50k scores and 30-day retention, then $0.50 per GB per monthStartups: The pricing page lists 6 to 12 months free for qualifying startups
Braintrust covers evals that compare prompts and models, logging and tracing, datasets, and Loop, an AI agent that analyzes data and builds evals. Starter is $0 per month with 1 GB processed data, 10k scores and 14-day retention; Pro is $249 per month with 5 GB, 50k scores and 30-day retention. Users, projects, datasets and experiments are unlimited on Starter, and qualifying startups get 6 to 12 months free.
Go ifAI product teams that run evals on every release and want unlimited users from the free tier up
Skip ifYou need long retention on a small budget
Why #3 over #4: It makes unlimited users free, but Pro at $249 per month costs more than Langfuse Pro at $199 per month.
4
Arize AX
Answrs VerdictHosted tracing and evals with an AI agent built in.
Headline feature: Tracing of inputs, outputs, tools and costs; LLM-as-a-judge, agent-as-a-judge and code evaluators; dashboards and alertsPrice: AX Free is $0 per month and AX Pro is $50 per month; AX Enterprise is customFree limits: 25k trace spans, 1 GB ingestion and 15 days retention per monthPro limits: 50k trace spans, 10 GB ingestion and 30 days retention per monthAI assistant: Alyx is the AI engineering agent built into Arize AX, and Signal surfaces recurring issues from production traces
Arize AX traces inputs, outputs, tools and costs, scores them with LLM-as-a-judge, agent-as-a-judge or code evaluators, and adds dashboards and alerts. AX Free is $0 with 25k spans, 1 GB and 15 days per month; AX Pro is $50 per month with 50k spans, 10 GB and 30 days. Alyx is the AI engineering agent built into the product and Signal surfaces recurring issues from production traces.
Go ifTeams that want a hosted platform with an in-product AI agent for debugging production traces
Skip ifYou want a published per-unit overage price
Why #4 over #5: Its $50 Pro plan is cheaper than Braintrust’s $249 Pro plan, though it includes 50k spans against Opik’s 100k at $19.
5
Weights & Biases Weave
Answrs VerdictBest fit when your team already lives in Weights & Biases.
Headline feature: Tracing with built-in agent SDK integrations or OpenTelemetry spans, plus evaluations with LLM judges and custom scorersPrice: Free is $0 per month; Pro starts at $60 per month billed monthly; Enterprise is customFree limits: 5 GB storage and 1 GB Weave data ingestion per monthPro limits: 100 GB storage and 1.5 GB Weave data ingestion per monthOverage: Additional ingestion is $0.10 per MB on Pro; additional storage is $0.03 per GB
Weave records agent and LLM app activity through built-in agent SDK integrations, OpenTelemetry spans or manual instrumentation, and runs evaluations with LLM judges and custom scorers. Free is $0 with 5 GB storage and 1 GB ingestion per month; Pro starts at $60 per month with 100 GB storage and 1.5 GB ingestion, and extra ingestion is $0.10 per MB. The pricing page now sits on CoreWeave’s site.
Go ifML teams already tracking experiments in Weights & Biases who want LLM tracing in the same account
Skip ifYou want pricing that is not metered by the megabyte
Why #5 over #6: It meters ingestion by the megabyte, which makes cost harder to predict than Datadog’s per-span pricing.
6
Datadog LLM Observability
Answrs VerdictThe pick for existing Datadog shops, with a $0 tier for 40K spans.
Headline feature: Tracing across prompts, retrieval, tool calls and agent decisions, with built-in or custom evaluators, experiments and versioned datasetsPrice: Free is $0 with up to 40K LLM spans per month; Pro starts at $160 per month on an annual commitmentPro volume: Pro includes 100K LLM spans per month; extra spans are $3.50 per 10K billed annuallyBilling unit: Priced per LLM span only, not tool, workflow or retrieval spansSecurity: Catches hallucinations, prompt injection attempts and PII exposure, with Sensitive Data Scanner built in
Datadog LLM Observability traces prompts, retrieval steps, tool calls and agent decisions, with built-in or custom evaluators, experiments and versioned datasets. Free covers up to 40K LLM spans per month with 15-day retention; Pro starts at $160 per month on an annual commitment with 100K spans, then $3.50 per 10K spans billed annually. Only LLM spans are billed, and Sensitive Data Scanner is included.
Go ifTeams already on Datadog who want LLM traces next to their APM, logs and infrastructure data
Skip ifYou do not run Datadog and want a standalone tool
Why #6 over #7: It prices only LLM spans, but Pro starts at $160 per month against $19 for Comet Opik.
7
Comet Opik
Answrs VerdictThe cheapest published paid cloud plan on this list at $19 per month.
Headline feature: Deep tracing of LLM calls, datasets, experiments and LLM-as-a-judge metrics, plus dashboards and online evaluation rulesPrice: Open Source and Free Cloud are $0; Pro Cloud is $19 per month; Enterprise is customFree Cloud: Up to 10 team members, 25k spans per month and 60 days of dataPro Cloud: Up to 50 team members and 100k spans per month; extra spans cost $5 per 100kLicense: Apache-2.0; the README says the full platform is free to self-host
Opik traces LLM calls, conversations and agent activity, with datasets, experiments, LLM-as-a-judge metrics for hallucination, moderation and RAG, plus dashboards and online evaluation rules. Open Source and Free Cloud are $0; Free Cloud allows 10 team members, 25k spans and 60 days. Pro Cloud is $19 per month for up to 50 team members and 100k spans, with extra spans at $5 per 100k. It is Apache-2.0 and free to self-host.
Go ifBudget-conscious teams that want the lowest published paid tier and a fully open-source option
Skip ifYou want long default retention without paying for an add-on
Why #7 over #8: It has the lowest paid plan on this list at $19 per month, but it ranks behind Langfuse because its documentation lists no prompt management.
8
MLflow
Answrs VerdictApache 2.0 and the broadest ML heritage on the list.
Headline feature: Tracing built on OpenTelemetry for any LLM provider and agent framework, evaluation with 50+ built-in metrics and LLM judges, prompt registry and an AI GatewayPrice: The GenAI page describes MLflow as 100% open source under the Apache 2.0 license and mentions no hosted pricingIntegrations: 100+ integrations and extensible APIsHosting: Databricks is linked as a managed MLflow option; MLflow also runs on Amazon SageMaker, Azure Machine Learning, Red Hat OpenShift AI and NebiusAdoption: The repository shows 28.6k GitHub stars
MLflow offers tracing built on OpenTelemetry for any LLM provider and agent framework, evaluation with 50+ built-in metrics and LLM judges, a prompt registry with lineage tracking and an AI Gateway. It is Apache 2.0, has 100+ integrations, and the repository shows 28.6k GitHub stars. Databricks is linked as a managed option, and MLflow also runs on Amazon SageMaker, Azure Machine Learning, Red Hat OpenShift AI and Nebius.
Go ifTeams that want one open-source tool for classic ML tracking and LLM tracing under a permissive license
Skip ifYou want a vendor-hosted product with a published price list
Why #8 over #9: It publishes no hosted price on the GenAI page, which makes it harder to budget than Opik or Langfuse.
9
Arize Phoenix
Answrs VerdictA free, local-first way to start with traces and evals.
Headline feature: Tracing, evaluation, prompt engineering, datasets and experiments, with a playground and a remote MCP serverPrice: Phoenix is open source and self-hosted; the pricing page lists no price for itLicense: Elastic License 2.0 (ELv2)Deployment: Runs on a local machine, in a container or in the cloud, with Docker Hub images and a Helm chart for KubernetesBuilt-in agent: PXI, a built-in AI agent for debugging traces and iterating on prompts
Phoenix covers tracing, evaluation, prompt engineering, datasets and experiments, with a playground, PXI (a built-in AI agent) and a remote MCP server. It runs on a laptop, in a container or in the cloud with Docker Hub images and a Helm chart. The license is Elastic License 2.0, and Arize’s pricing page lists no price for Phoenix itself.
Go ifDevelopers who want to start tracing locally in minutes before committing to a hosted platform
Skip ifYour policy requires an OSI-approved license
Why #9 over #10: It is source-available under ELv2 rather than MIT or Apache, which is why it ranks below Langfuse and MLflow.
10
New Relic AI Monitoring
Answrs VerdictAPM-first, so it is strongest as an add-on to New Relic.
Headline feature: A summary of AI metrics from response time and quality to tokens, APM golden signals and infrastructureAgents: Maps every interaction across multi-agent systems and shows which tools are called by which AI agentsPrice: Usage-based: 100 GB per month of data ingest free, then $0.40 per GB, with one free full platform userAI pricing: The pricing page lists no separate price for AI monitoringIntegrations: Built-in integrations for OpenAI, Pinecone and LangChain, with OpenTelemetry as the instrumentation provider
New Relic AI Monitoring shows AI metrics from response time and quality to tokens, APM golden signals and infrastructure, maps interactions across multi-agent systems and monitors the Model Context Protocol request lifecycle. It has built-in integrations for OpenAI, Pinecone and LangChain and uses OpenTelemetry. Pricing is usage based: 100 GB per month of ingest free, then $0.40 per GB, with no separate AI monitoring price listed.
Go ifTeams already on New Relic who want AI metrics next to application performance data
Skip ifYou want evals, datasets and prompt versioning at the center of the tool
How we ranked these
We ranked platforms on how many core LLMOps jobs each vendor documents on its own pages (tracing, evaluation, prompt management), whether a free or open-source option exists, and how clearly the price is published. Every price is from the vendor’s pricing page the day we checked, and where a page did not state a figure we say so. We left out Humanloop, Galileo and Portkey: Galileo’s pricing page now redirects to Splunk, Portkey has been acquired by Palo Alto Networks, and Humanloop’s own site says it is sunsetting the platform as its team joins Anthropic, so we did not rank them as standalone picks. Helicone’s pricing page states it is no longer part of Mintlify with details to come, so we left it out too. Facts are taken from each maker's own published information; check the official site before you buy.
Questions people ask
What is the best LLMOps platform?
Langfuse is our top pick: tracing, prompt management and evaluation in one tool, open source under MIT apart from the ee folders, with a free Hobby plan of 50k units per month. LangSmith is #2 and Braintrust is #3.
Which LLMOps platforms are open source?
Langfuse is MIT licensed except for its ee folders, Comet Opik is Apache-2.0, MLflow is Apache 2.0, and Arize Phoenix is under the Elastic License 2.0, which is source-available rather than an OSI-approved license.
What is the cheapest paid LLMOps plan?
Comet Opik Pro Cloud at $19 per month with 100k spans, then Langfuse Core at $29 per month, LangSmith Plus at $39 per seat per month and Arize AX Pro at $50 per month, per each vendor’s pricing page.
Which options sit inside an existing observability platform?
Datadog LLM Observability and New Relic AI Monitoring sit inside the vendors’ existing observability platforms. Datadog Free covers up to 40K LLM spans per month, and New Relic includes 100 GB per month of free data ingest.
Advertised monthly price of the lowest paid plan, in US dollars (LangSmith is per seat; Datadog is billed annually; usage overage excluded) at a glance
Comet Opik19 USD per month
Langfuse29 USD per month
LangSmith39 USD per month
Arize AX50 USD per month
Weights & Biases Weave60 USD per month
Datadog LLM Observability160 USD per month
Braintrust249 USD per month
Advertised monthly price of the lowest paid plan, in US dollars (LangSmith is per seat; Datadog is billed annually; usage overage excluded), as published by each maker. Picks without a published figure are left off.
