Comparison · Updated Oct 10, 2026 · published Aug 27, 2026, revised 2 times
Arize AI Alternatives (2026): Langfuse, LangSmith, Fiddler, and When to Stay
Alternatives to Arize AI depend on the job. Langfuse for open-source and many users. LangSmith for LangChain deployment and evals. Fiddler for guardrails priced per trace. Helicone only if you are already on it.
Pricing and capability claims verified through .
Moneda may earn a commission from some product links. Partner economics do not affect our recommendations. This page is research-based — conclusions come from product evidence and documentation, not a claim that we personally purchased every product listed.
Which of these are you?
Your pick
Langfuse
Langfuse Core is $29 a month with unlimited users and 100k units included. Self-hosting the open-source build is free. A unit is a trace, an observation, or a score — not a seat.
Tradeoff
Leave Arize for Langfuse when you want open-source, framework-agnostic tracing and a bill that is not per seat. Leave it for LangSmith when the stack is LangChain or LangGraph and you want deployment plus evals. Leave it for Fiddler when the job is guardrails at a published per-trace rate. Stay on Helicone only if you are already there — it is in maintenance mode, and it is not the default alternative.
| Feature | Arize Self-hostedFrom $50/moEnterprise | Langfuse Self-hostedFrom $8/moEnterprise | LangSmith Self-hostedFrom $1/moEnterprise | Fiddler Free tierFrom $0/moEnterprise | Helicone From $79/moEnterprise |
|---|---|---|---|---|---|
| Pricing | Self-hostedFrom $50/moEnterprise AX Free: $0, 25k spans/month, 1 GB, 15-day retention, unlimited users - Checked 2026-10-10: https://arize.com/pricing/ · AX Pro: $50/month, 50k spans/month, 10 GB, 30-day retention, unlimited users - Checked 2026-10-10: https://arize.com/pricing/ · AX Enterprise: custom, SaaS or self-hosted - Checked 2026-10-10: https://arize.com/pricing/ · Phoenix: open source, free to use, self-hosted - Checked 2026-10-10: https://arize.com/docs/phoenix/resources/frequently-asked-questions/what-is-the-difference-between-phoenix-and-arize | Self-hostedFrom $8/moEnterprise Hobby: $0, 50k units/month, 30-day data access, 2 users - Checked 2026-10-10: https://langfuse.com/pricing · Core: $29/month, 100k units included then $8/100k, unlimited users, 90-day data access - Checked 2026-10-10: https://langfuse.com/pricing · Pro: $199/month, 100k units included, 3-year data access - Checked 2026-10-10: https://langfuse.com/pricing · Enterprise: $2,499/month - Checked 2026-10-10: https://langfuse.com/pricing · Self-hosted open source: free under the MIT license, no usage bill - Checked 2026-10-10: https://langfuse.com/docs/administration/billable-units | Self-hostedFrom $1/moEnterprise Developer: $0/seat, 1 seat, 5,000 base traces/month, then 0.005 LSU per extra base trace (1 LSU = $1) - Checked 2026-10-10: https://www.langchain.com/pricing · Plus: $39/seat/month, 10,000 base traces included, then 0.005 LSU per extra base trace - Checked 2026-10-10: https://www.langchain.com/pricing · Extended retention upgrade: 0.0025 LSU per trace. Base traces kept 14 days, extended traces 180 days - Checked 2026-10-10: https://www.langchain.com/pricing · Enterprise: custom. Cloud, hybrid, or self-hosted - Checked 2026-10-10: https://www.langchain.com/pricing | Free tierFrom $0/moEnterprise Free: guardrails listed, no dollar price and no included-trace count on the plan card - Checked 2026-10-10: https://www.fiddler.ai/pricing · Developer: $0.002 per trace - Checked 2026-10-10: https://www.fiddler.ai/pricing · Enterprise: custom, SaaS, VPC, or on-prem - Checked 2026-10-10: https://www.fiddler.ai/pricing | From $79/moEnterprise Hobby: 10,000 requests/month, 1 GB storage, 1 seat, 1 organization, 7-day retention. The page calls the requests free and does not print a Hobby dollar price - Checked 2026-10-10: https://www.helicone.ai/pricing · Pro: $79/month, unlimited seats, 1-month retention, usage-based past the included requests and storage - Checked 2026-10-10: https://www.helicone.ai/pricing · Team: $799/month, 5 organizations, 3-month retention, SOC-2 and HIPAA listed - Checked 2026-10-10: https://www.helicone.ai/pricing · Enterprise: on-prem deployment, retention listed as forever - Checked 2026-10-10: https://www.helicone.ai/pricing |
| Free tier | ❌ | ❌ | ❌ | ✅ | ❌ |
| Open-source and self-host path | ✅ | ✅ | ⚠️ | ⚠️ | ✅ |
| LangChain deployment and evals | ⚠️ | ⚠️ | ✅ | ❌ | ❌ |
| How the bill is metered | ✅ | ✅ | ✅ | ✅ | ⚠️ |
| Best for | Teams running production AI systems that need model and LLM observability | Teams that want open-source or self-hosted tracing, or a cloud plan that does not charge per seat | Teams building on LangChain who need tracing and evals | Enterprises needing comprehensive AI observability, explainability, and governance with flexible deployment options. | Teams already running Helicone, or teams that specifically want its self-hosted proxy |
| Direct link | View Arize | View Langfuse | View LangSmith | View Fiddler | View Helicone |
✅ Included · ❌ Missing · ⚠️ Partial or add-on
Arize
AX Free: $0, 25k spans/month, 1 GB, 15-day retention, unlimited users - Checked 2026-10-10: https://arize.com/pricing/ · AX Pro: $50/month, 50k spans/month, 10 GB, 30-day retention, unlimited users - Checked 2026-10-10: https://arize.com/pricing/ · AX Enterprise: custom, SaaS or self-hosted - Checked 2026-10-10: https://arize.com/pricing/ · Phoenix: open source, free to use, self-hosted - Checked 2026-10-10: https://arize.com/docs/phoenix/resources/frequently-asked-questions/what-is-the-difference-between-phoenix-and-arize
Best for: Teams running production AI systems that need model and LLM observability
- Free tier
- ❌
- Open-source and self-host path
- ✅
- LangChain deployment and evals
- ⚠️
- How the bill is metered
- ✅
Langfuse
Hobby: $0, 50k units/month, 30-day data access, 2 users - Checked 2026-10-10: https://langfuse.com/pricing · Core: $29/month, 100k units included then $8/100k, unlimited users, 90-day data access - Checked 2026-10-10: https://langfuse.com/pricing · Pro: $199/month, 100k units included, 3-year data access - Checked 2026-10-10: https://langfuse.com/pricing · Enterprise: $2,499/month - Checked 2026-10-10: https://langfuse.com/pricing · Self-hosted open source: free under the MIT license, no usage bill - Checked 2026-10-10: https://langfuse.com/docs/administration/billable-units
Best for: Teams that want open-source or self-hosted tracing, or a cloud plan that does not charge per seat
- Free tier
- ❌
- Open-source and self-host path
- ✅
- LangChain deployment and evals
- ⚠️
- How the bill is metered
- ✅
LangSmith
Developer: $0/seat, 1 seat, 5,000 base traces/month, then 0.005 LSU per extra base trace (1 LSU = $1) - Checked 2026-10-10: https://www.langchain.com/pricing · Plus: $39/seat/month, 10,000 base traces included, then 0.005 LSU per extra base trace - Checked 2026-10-10: https://www.langchain.com/pricing · Extended retention upgrade: 0.0025 LSU per trace. Base traces kept 14 days, extended traces 180 days - Checked 2026-10-10: https://www.langchain.com/pricing · Enterprise: custom. Cloud, hybrid, or self-hosted - Checked 2026-10-10: https://www.langchain.com/pricing
Best for: Teams building on LangChain who need tracing and evals
- Free tier
- ❌
- Open-source and self-host path
- ⚠️
- LangChain deployment and evals
- ✅
- How the bill is metered
- ✅
Fiddler
Free: guardrails listed, no dollar price and no included-trace count on the plan card - Checked 2026-10-10: https://www.fiddler.ai/pricing · Developer: $0.002 per trace - Checked 2026-10-10: https://www.fiddler.ai/pricing · Enterprise: custom, SaaS, VPC, or on-prem - Checked 2026-10-10: https://www.fiddler.ai/pricing
Best for: Enterprises needing comprehensive AI observability, explainability, and governance with flexible deployment options.
- Free tier
- ✅
- Open-source and self-host path
- ⚠️
- LangChain deployment and evals
- ❌
- How the bill is metered
- ✅
Helicone
Hobby: 10,000 requests/month, 1 GB storage, 1 seat, 1 organization, 7-day retention. The page calls the requests free and does not print a Hobby dollar price - Checked 2026-10-10: https://www.helicone.ai/pricing · Pro: $79/month, unlimited seats, 1-month retention, usage-based past the included requests and storage - Checked 2026-10-10: https://www.helicone.ai/pricing · Team: $799/month, 5 organizations, 3-month retention, SOC-2 and HIPAA listed - Checked 2026-10-10: https://www.helicone.ai/pricing · Enterprise: on-prem deployment, retention listed as forever - Checked 2026-10-10: https://www.helicone.ai/pricing
Best for: Teams already running Helicone, or teams that specifically want its self-hosted proxy
- Free tier
- ❌
- Open-source and self-host path
- ✅
- LangChain deployment and evals
- ❌
- How the bill is metered
- ⚠️
What we evaluated
Evaluation criteria: Open-source and self-host path, LangChain deployment and evals, How the bill is metered. Conclusions are based on primary documentation, product evidence, and workflow fit — not hands-on benchmark runs unless noted. 8 primary sources checked.
Arize is still a reasonable place to stay. AX Free is $0 for 25,000 spans a month, AX Pro is $50 a month for 50,000 spans, and both plans list unlimited users. Phoenix is the open-source product you run yourself. This page is for the job Arize is not: open-source cloud with unit billing, LangChain deployment, or guardrails priced per trace.
Helicone is not the first alternative. Mintlify acquired Helicone on 3 March 2026. Helicone’s letter the same day says the service remains live in maintenance mode — security updates, new models, and bug and performance fixes. Use it if you already run it. Do not start a new team there. The head-to-head is Helicone vs Arize.
Quick comparison
Prices checked 10 October 2026 on each vendor’s pricing page.
| Tool | Best for | What you can price | Fit |
|---|---|---|---|
| Arize | Stay, if spans and Phoenix fit | AX Free $0 (25k spans). AX Pro $50/month (50k spans). Unlimited users. Phoenix free to self-host | Incumbent |
| Langfuse | Open source, many users, no seat fee | Hobby 50k units. Core $29/month. A unit is a trace, observation, or score | Default alternative |
| LangSmith | LangChain/LangGraph plus deployment | Developer $0, 1 seat, 5k base traces. Plus $39/seat, 10k base traces, then 0.005 LSU per extra base trace ($0.005) | LangChain alternative |
| Fiddler | Guardrails and a per-trace rate | Developer $0.002 per trace. Enterprise is custom | Guardrails alternative |
| Helicone | Already deployed | Hobby includes 10k requests. Pro $79/month. Team $799/month. Maintenance mode | Existing installs only |
When to stay on Arize
Stay when the operating model is the point. Phoenix if your team will run the open-source product. AX if you want Arize to run SaaS and the included span pool is enough. The pricing page does not print a dollar rate for spans past the included amount, so a workload above 50,000 spans a month on Pro needs a sales conversation or a move to another meter you can calculate.
When not to stay
Move if you need LangSmith Deployment for LangGraph, if you want Langfuse’s unit price and unlimited users on a $29 plan, or if Fiddler’s per-trace guardrail plan matches the buy. Do not move to Helicone for a lighter setup. That recommendation is withdrawn.
Langfuse
Langfuse is the default alternative for a team that is not all-in on LangChain. Cloud plans are workspace fees plus units. Core is $29 a month, unlimited users, 100,000 units included, then $8 per 100,000 units until higher volume tiers. Self-hosting the open-source build is free under the MIT license. The head-to-head on billing units versus LangSmith traces is Langfuse vs LangSmith.
LangSmith
LangSmith is the alternative when the application is LangChain or LangGraph and you want tracing, evals, and deployment together. Plus is $39 per seat per month with 10,000 base traces included for the organization, then pay-as-you-go. Developer is one free seat and 5,000 base traces. Base traces are kept 14 days. Extended traces are kept 180 days and add 0.0025 LSU each. Details and a worked example are on LangSmith vs Arize.
Fiddler
Fiddler’s pricing page lists a Free guardrail plan, Developer at $0.002 per trace, and Enterprise for SaaS, VPC, or on-prem. Choose it when explainability, fairness, or those guardrails are the buying requirement. The three-way write-up is Arize vs Fiddler vs WhyLabs.
Helicone
Choose Helicone only when the proxy is already in production, or you have already decided to run Helicone’s documented self-host. The 3 March 2026 letter is the constraint: maintenance mode, not a new feature line. Pro is $79 a month and Team is $799 a month on helicone.ai/pricing. Both say usage-based pricing applies. The plan table does not state a per-request dollar rate, so this page does not invent one.
A short cost picture
Five people, 50,000 agent runs in a month, 10 spans inside each run, no extra scores. This uses each vendor’s own unit. It is an illustration, not a quote.
- LangSmith Plus: 5 × $39 = $195 in seats. 10,000 base traces are included. The other 40,000 cost 0.005 LSU each. One LSU is $1, so the trace line is $200. Seats plus base traces are $395, before extended retention or deployment.
- Langfuse Core: those runs are 50,000 traces plus 500,000 observations = 550,000 units. Core is $29, the first 100,000 units are included, and the next 450,000 are $8 per 100,000, which is $36. About $65. Users on Core are unlimited, so five people do not add seats.
- Arize AX Pro: $50 for unlimited users and 50,000 spans included. Ten spans times 50,000 runs is 500,000 spans. That is past the included pool. Arize does not print the overage rate on the plan card.
- Fiddler Developer: $0.002 × 50,000 traces = $100, if each run is one trace. The page does not list a seat fee on Developer.
- Helicone Pro: $79 and 10,000 included requests, then a usage charge that is not printed as a per-request rate. Not the plan to start on.
What we evaluated: self-host and open-source paths, LangChain deployment, and the meter on the invoice. Langfuse won for a new framework-agnostic team. LangSmith won for LangChain deployment. Fiddler won for a published per-trace guardrail rate. Helicone won only for an existing install.
An earlier version of this page treated WhyLabs as an active option and put Helicone’s button first. WhyLabs, Inc. was verified on 2026-08-30 (https://whylabs.ai/) to be discontinuing operations. Helicone’s maintenance-mode status was verified on 10 October 2026. Neither is the recommended start for a new buyer.