Arize Pricing
Arize costs $0 to start and $50 a month on Pro. Its own pricing page never mentions the free tier that comes before either one.
Arize's real numbers, verified straight from its own pricing page: AX Free costs $0 a month, AX Pro costs $50 a month, and AX Enterprise has no published price at all, quoted only through sales. What that page leaves out is Phoenix, Arize's own open source tracing and evaluation tool, genuinely free and self-hostable with no span cap at all, sitting in Arize's own docs but never shown on the same pricing table as the three paid tiers. Every AX tier also bills against a span, a unit Arize never defines in plain language anywhere on its site, and a $60,000-a-year Enterprise figure circulates on other pricing writeups that Arize's own page has never published. This page verifies every number, adds the tier Arize leaves off its own table, and covers the one question none of the four answer: what does any of it show the customer paying for the agent underneath it.
Verified against Arize's own pricing page
What Arize actually costs
Arize's AX pricing runs three published tiers, and every figure below comes straight from Arize's own current pricing page. AX Free costs $0 a month, includes twenty five thousand spans and one gigabyte of ingestion each month, holds fifteen days of retention, and still ships with unlimited users, unlimited evaluations, and unlimited experiments even at that price, plus ten Signal issues a month for its automated debugging feature. AX Pro costs $50 a month, doubles the span allowance to fifty thousand and the ingestion allowance to ten gigabytes, stretches retention to thirty days, raises Signal issues to twenty five a month, and adds email support in place of Free's community-only channel. Arize's own docs also list a thirteen month extended retention add-on that a Pro workspace can buy without moving to a different base plan, a detail that never shows up on the main pricing page itself.
AX Enterprise is where the published numbers stop. Arize's page lists span volume, storage, and retention all as custom, reachable only through a sales conversation, and adds the features the two paid-by-the-month tiers don't get: self-hosted deployment on top of the standard SaaS option, enterprise SSO, audit logs, HIPAA compliance, managed agents, and access to Arize's Data Fabric for larger-scale telemetry. What Arize's own pricing page never shows on that same table is Phoenix, its separate open source tracing and evaluation project, genuinely free under an open license, self-hostable with no span cap, no ingestion cap, and no retention window at all, since a self-hosted deployment keeps whatever a team's own infrastructure can hold. Phoenix lives in Arize's docs and on its own product pages, not on the pricing table next to Free, Pro, and Enterprise, so a reader comparing Arize pricing against three numbers is actually choosing between four real options without realizing a fourth one exists.
Four options, not three
Arize's real structure: four options, not three
Straight from Arize's own pricing page and docs, not a third party estimate.
Phoenix: genuinely free
Open source tracing, evaluation, prompt engineering, and dataset tooling, self-hosted with no span cap and no retention limit. The one option Arize's own pricing page never lists next to the other three.
AX Free: $0 a month
Twenty five thousand spans, one gigabyte of ingestion, fifteen days of retention. Unlimited users, evaluations, and experiments even at this tier.
AX Pro: $50 a month
Fifty thousand spans, ten gigabytes of ingestion, thirty days of retention, email support, and a thirteen month retention add-on Arize's docs mention but its pricing page doesn't.
AX Enterprise: custom pricing
Custom span volume, storage, and retention, self-hosted or SaaS, enterprise SSO, audit logs, HIPAA, managed agents, and Data Fabric access. No published number anywhere.
The number every tier bills against, undefined
What actually counts as one span?
Every AX tier bills against a span, and Arize's own pricing page and its usage documentation both use the word on every row without ever defining what one span actually is. Tracing tools generally treat a span as one recorded step inside a trace, one model call, one tool call, one retrieval step, but Arize never states that mapping outright anywhere public, so a team pricing out a multi-step agent has to estimate its own span count from a real workload rather than read a definition off the page. Ingestion, the second meter every tier bills against alongside spans, fares slightly better: Arize's docs at least name it as raw data volume in gigabytes, though how a single trace's size in gigabytes relates to its span count is left for the reader to work out from their own account once real traffic starts flowing through it.
What happens once a workspace goes past its included span or ingestion allowance is even less documented. Neither Arize's pricing page nor its usage docs describe an overage rate, a billed rate per extra span, or any mechanism for what happens after the limit hits, a gap that stands out against Braintrust, Langfuse, and PromptLayer, three comparable tools that each publish a specific per-unit overage rate on their own pricing pages. A third party pricing writeup on Arize describes hard caps that simply stop data collection once a tier's allowance runs out rather than billing an overage, but that behavior doesn't appear anywhere on Arize's own site, so treat it as an informed guess rather than a confirmed policy until Arize's own team confirms it against a real account.
The tier that changes where the data lives and who can prove it's secure
What only Enterprise unlocks
Free and Pro both run as shared SaaS with community or email support only.
Self-hosted deployment
Free and Pro both run only as Arize's own shared SaaS. Enterprise adds the option to self-host, priced through a sales conversation rather than a published rate.
SSO, audit logs, and HIPAA
None of Free or Pro's rows include single sign-on, audit logging, or HIPAA compliance. All three move to Enterprise only, alongside managed agents and Data Fabric access for larger-scale telemetry.
What the meter doesn't show
Arize's spans and gigabytes tell you what the agent did. They tell your customer nothing
Every number on Arize's pricing page, the span count climbing toward a plan's allowance, the ingestion gigabytes filling up a retention window, the Signal issues Arize's own debugging assistant surfaces, describes what happened inside your own account and what it cost your own team to trace it. That holds at every tier, including Enterprise, since even its enterprise SSO and audit logs still govern who inside your company can see or ship what, not whether an outside customer can see anything at all. Unlimited users on every AX tier, Free through Enterprise, means unlimited seats for your own engineers, and Phoenix's self-hosted, uncapped tracing changes where the data physically lives without changing who the dashboard was built for: the same engineers who had access on shared SaaS still have access on a self-hosted Phoenix instance, just on infrastructure their own company runs.
If you sell an AI agent inside a product other companies pay for, that gap shows up the moment a customer asks what the agent actually did for their account this month, a question no span count or ingestion total answers on its own. AiAgRe's Node SDK reads the same kind of underlying agent events Arize already traces, through the same LangChain, LlamaIndex, or CrewAI integrations, and tags each one with an organization identity and a customer identity at the point of ingestion. That turns into white-labeled dashboard components your own customer sees inside your product, showing deflection rate and cost per resolution scoped so tightly that one customer's numbers never reach another's view. It doesn't replace Arize's own tracing or Phoenix's evaluation tooling: most teams run both, Arize or Phoenix for what happened inside a trace, AiAgRe for what that same agent was worth to the person paying for it.
When a span count stops being the real question
Three questions Arize's pricing page was never built to answer
None of these show up in a span count, a gigabyte total, or a Signal issue. All three show up once real customers start asking.
Your customer asks what their account got, not how many spans ran
A customer deciding whether to renew wants their own deflection rate and cost per resolution, scoped to their own traffic, not a walkthrough of a workspace they can't log into. A clean trace reassures your engineers; it says nothing to the person deciding whether to keep paying.
Unlimited seats for your team, zero seats for theirs
Every AX tier, Free through Enterprise, removes the meaningful limits as a workspace scales: more spans, more gigabytes, more retention. None of it adds a scoped view for a customer outside your company, because Arize's workspace model was built for your own engineers, not for a second tenant layer underneath them.
The span count tracks volume. Nothing tracks value
Arize's meters answer what an agent's traffic costs to trace as it grows, on a unit Arize's own page never fully defines. They say nothing about what that traffic was worth, the same gap a fuller look at AI agent evaluation tools as a category runs into past Arize alone.
FAQs
Arize pricing: frequently asked questions
Common questions from teams pricing Arize against their own usage before they commit to a tier, or before they add a second layer on top.
How much does Arize actually cost?
Three published tiers, verified straight from Arize's own pricing page, plus a fourth option that page never lists next to them. AX Free costs $0 a month for twenty five thousand spans, one gigabyte of ingestion, and fifteen days of retention. AX Pro costs $50 a month, doubling spans to fifty thousand, ingestion to ten gigabytes, and retention to thirty days. AX Enterprise has no published number at all, quoted only through Arize's sales team, and adds self-hosting, enterprise SSO, audit logs, and HIPAA compliance that neither paid-by-the-month tier includes. The fourth option is Phoenix, Arize's own open source tracing and evaluation tool, genuinely free and self-hostable with no span cap, sitting in Arize's docs rather than on the same pricing table as the other three.
What's actually included on the AX Free plan, and how far does it really go?
Twenty five thousand spans, one gigabyte of ingestion, and fifteen days of retention a month, alongside unlimited users, unlimited evaluations, and unlimited experiments even at the free tier, plus ten Signal issues a month through Arize's automated debugging feature, all support handled at community level with no dedicated channel. The span allowance is usually what a team hits first, not the user count: a five person team tracing a moderately busy agent can burn through twenty five thousand spans well before it runs into any seat limit, since Free never caps how many people can see the account, only how much the account can trace.
What is a 'span', and why doesn't Arize define it anywhere?
Every AX tier bills against spans, and neither Arize's pricing page nor its usage documentation ever states in plain language what counts as one. Most tracing tools treat a span as one recorded step inside a trace, one model call, one tool call, one retrieval step, and Arize's own product behavior suggests the same reading, but Arize never confirms that mapping outright on any public page. A team pricing out a multi-step agent has no published formula to work from and has to estimate its real span count from an actual workload rather than read a definition off the pricing table.
What happens once a workspace goes over its span or ingestion limit?
Arize's own pricing page and usage docs both leave this undocumented: no overage rate, no per-span billed rate, and no stated mechanism for what happens after a tier's allowance runs out, a gap that stands out next to Braintrust, Langfuse, and PromptLayer, three comparable tools that each publish a specific overage rate on their own pricing pages. One third party pricing writeup on Arize describes a hard cap that stops data collection once the limit hits rather than billing an overage, but that behavior isn't confirmed anywhere on Arize's own site, so treat it as an informed guess rather than a stated policy until Arize's own team confirms it against a real account.
Is Phoenix really free, and how is it different from AX Free?
Yes, and it's a different product from AX Free rather than a smaller version of it. Phoenix is Arize's open source tracing, evaluation, prompt engineering, and dataset tool, self-hostable with no span cap, no ingestion cap, and no retention window built into its pricing at all, since a self-hosted deployment keeps whatever a team's own infrastructure can hold. AX Free is Arize's own hosted SaaS tier, capped at twenty five thousand spans and fifteen days of retention, with Arize managing the infrastructure behind it. A team with the engineering time to run its own deployment gets a genuinely uncapped tool for free on Phoenix; a team that wants Arize to host it trades that cap for zero setup on AX Free.
Does Arize Enterprise really cost $60,000 a year?
That figure shows up on at least one third party pricing writeup as a median deal size, but it doesn't appear anywhere on Arize's own site, which lists Enterprise only as custom pricing reachable through a sales conversation, with no minimum contract value or starting number published anywhere. What Arize's page does document is the feature set added at that tier: self-hosted deployment on top of the standard SaaS option, enterprise SSO, audit logs, HIPAA compliance, managed agents, and access to Arize's Data Fabric for larger-scale telemetry. Treat any specific Enterprise dollar figure you see on a third party page as an estimate, not a confirmed number, until Arize's own sales team quotes one against your real usage.
Is Arize cheaper than Langfuse, LangSmith, or Braintrust?
At the paid entry tier, yes on sticker price: Arize's AX Pro runs $50 a month against Langfuse's Pro tier at $199, LangSmith's Plus plan at $39 a seat a month, and Braintrust's Pro tier at $249 a month. All four meter completely different units though, spans and gigabytes for Arize, traces and observations for Langfuse, traces plus LangChain compute and storage units for LangSmith, data and scores for Braintrust, so a straight dollar comparison undercounts what each one actually measures. The fuller side by side, including where Confident AI fits into the same comparison, lives on the LangSmith alternatives page, and the full unit-by-unit Langfuse breakdown lives on the Langfuse pricing page.
Does any Arize plan show pricing or ROI data to my own customers, and can I run AiAgRe alongside it?
No plan does, AX Free through Enterprise, Phoenix included, and yes, running both is the normal setup rather than a workaround. Every dashboard Arize renders, the span count climbing toward a plan's allowance, the ingestion gigabytes filling a retention window, Signal's automated issues, lives inside a workspace built for the team that owns it. Unlimited users on every AX tier means unlimited seats for your own engineers, not a scoped view for an outside customer, and even Enterprise's SSO and audit logs govern who inside your company sees what, never a second tenant layer underneath it. AiAgRe's Node SDK reads the same kind of underlying agent events Arize already traces through its own tracing, tags each one with an organization identity and a customer identity at the point of ingestion, and turns that into white-labeled dashboard components your own customer sees inside your product, showing deflection rate and cost per resolution scoped so tightly that one customer's numbers never reach another's view. Installing it doesn't touch how Arize counts a span or an ingestion gigabyte; the two SDKs read the same agent, one billed and displayed to your team, the other billed and displayed to the customer paying for it.
Already tracing your agent with Arize? Now show your customer what it did.
Request access and we'll walk through how AiAgRe's embed tokens map onto the trace data Arize already produces.