Pricing

Simple pricing for complex AI systems.

Your first month

Your first $950 of usage is on us.

Sign up, connect your cloud and run real workloads. No card at signup.

Enterprise
Custom
Annual commit, fully credited to usage. Significant discounts on every meter.

For organizations running mission-critical AI, ML and agents across teams, regions and clouds.

  • Significant volume discounts on every meter
  • Unlimited clusters, multi-region and multi-cloud
  • Custom concurrency, fanout and retention
  • Zero Trust security and observability
  • SSO and SAML, custom RBAC
  • Self-Hosted deployment option available
  • Premier support, 24×7
  • Observability

“Union.ai has been critical at the time that we needed to significantly scale up… Union.ai’s wealth of expertise has enabled us to move fast and deliver at scale.”

Alborz Alavian, Senior Engineering Manager, Woven by Toyota
Rate card

Four meters. Published ladders.

You pay for what each action consumes: the action itself, plus the vCPU, memory and GPU it allocates, metered to the second. Every ladder is measured on the volume you use in a billing cycle and starts again from the first band when the next cycle begins. Nothing carries over.

Enterprise

Enterprise plans are priced on an annual commit and carry significant volume discounts on every meter: actions, vCPU, memory and GPU.

Talk to sales

Actions

per action

One durable unit of work. A task, a trace, a condition, a cache hit, an artifact version or a tracked run each count as one, however long it runs and however many times it retries.

0 – 100k actions / mo
$0.0030
100k – 1M
$0.0015
1M – 10M
$0.0008
10M +
$0.0003

vCPU

per vCPU-hour

Allocated vCPU on every container running your actions, read from the task’s resource request and billed to the second.

0 – 100k vCPU-hrs / mo
$0.0230
100k – 1M
$0.0173
1M +
$0.0130

Memory

per GB-hour

Allocated memory, same basis as vCPU. Bands are six times the vCPU bands, so a normally shaped workload crosses both meters together.

0 – 600k GB-hrs / mo
$0.0032
600k – 6M
$0.0024
6M +
$0.0018

How the ladders work

example

Bands are marginal. A cycle with 2.5M actions is priced as three slices, not one rate:

first 100k × $0.0030
$300
next 900k × $0.0015
$1,350
next 1.5M × $0.0008
$1,200
2.5M actions, before compute
$2,850
Crossing a band reprices only the units above it, never the whole cycle.

GPU

per GPU-hour · by class · same on every cloud

Every accelerator belongs to one of three classes, set by the median on-demand price of that silicon across the clouds that sell it. The class rate is the same on AWS, GCP, Azure and neoclouds, and hours pool by class before the ladder runs.

Class0 – 1k hrs / mo1k – 10k10k +
Class 1 Accelerators under $1.50 / hr on the open market T4L4Inferentia2TPU v5eTrainium1 $0.068$0.056$0.045
Class 2 $1.50 – $5.50 / hr A10GL40SV100A100RTX PRO 6000Trainium2TPU v6eTPU v5p $0.24$0.20$0.16
Class 3 Over $5.50 / hr H100H200B200B300 $0.89$0.74$0.59
Reserved or owned GPUs Capacity reservations, committed-use contracts, or your own silicon. A flat monthly fee per GPU, unlimited hours. Talk to sales
$500per Union-managed cluster / month

Let Union run your infrastructure, optional. Union hosts the control plane on every plan. If you also want Union to provision, upgrade and operate the clusters in your cloud, it is $500 per cluster per month and never more than $2,500 a month. Self-managed, where you run the clusters yourself, carries no fee. New accounts get their first Union-managed cluster free in month one.

Compare plans
Start to scale

Start with Flyte. Graduate when you’re ready.

Union is built by the creators of Flyte. Team is the fastest way to start with Flyte, even at small scale. It costs little, sets up quickly and runs in your own cloud, so you keep your negotiated rates and your cloud credits. Teams that start on open-source Flyte move to Team when their workloads need more speed, scale, security or cost control, and to Enterprise when they need multiple clouds, custom security, single tenancy or committed-spend pricing.

See the three steps
Step 1
FlyteOpen source, run by you
FreeApache 2.0
What you get
The full Flyte SDK and engine: durable tasks, caching, retries, type-safe data, the Flyte UI.
Who runs it
You. Install and operate Flyte on your own Kubernetes.
Clusters
As many as you operate
Support
Community, on Slack and GitHub
Time to graduate when
  • Workloads need lower latency, faster execution, higher scale or higher availability
  • Workloads grow beyond one cluster
  • You need RBAC, stronger security and more automation, such as artifact triggers, or want to serve applications behind RBAC
  • You want to cut operating and compute cost. Union runs the control plane, and its engine runs outside Kubernetes, so the same cluster carries far more work. Usage and metrics, reusable containers, reusable Ray clusters and container cold boot cut the rest. Cold boot is coming soon to self-managed deployments.
  • You run production LLM endpoints that scale from zero, and need GPU observability, queues for team quotas and usage, and concurrency controls
Read the Flyte docs
Step 2
TeamManaged control plane, your cloud
$950 / mo minimumCredited to usage. First month waived.
What you get
Everything in Flyte, plus the Union control plane: Zero Trust, remote debugging, replay and recover, Volumes, federated login with SCIM, and published usage pricing. Bring your Flyte code as it is.
Who runs it
Union hosts the control plane. Run your clusters yourself at no fee, or have Union run them for $500 per cluster.
Clusters
2 active, 10,000 concurrent actions, 200,000 actions per run
Support
Standard, 12×5, Sev 1 in 4 business hours
Time to graduate when
  • You run across multiple clouds and regions on one control plane
  • Security asks for a private endpoint, CMEK, BYOK, your own secrets backend or custom RBAC
  • You need more scale than Team’s 2 clusters, 10,000 concurrent actions and 200,000 actions per run
  • You need a single-tenant control plane, or want to self-host
  • You need 24×7 support or forward-deployed engineers
  • You want lower unit rates in return for committed spend
Start with $950 of usage free
Step 3
EnterpriseEvery team, region and cloud
CustomAnnual commit, credited to usage
What you get
Everything in Team, plus significant discounts on every meter, private networking, CMEK and BYOK, SSO and SAML, custom RBAC and namespace topology, custom cloud configuration.
Who runs it
Union, in your cloud, on-prem or air-gapped. Forward-deployed engineers available.
Clusters
Unlimited, multi-region and multi-cloud on one control plane. Custom concurrency and fanout.
Support
Premier included, 24×7 for Sev 1–2. Mission Critical available as a paid upgrade.
You’re here when
  • AI is on the critical path for more than one team
  • Procurement wants one commit, drawn down by usage, across every environment
Talk to sales
Compare plans

What each plan includes

FeatureTeamEnterprise
Commercial
How you pay$950 monthly minimum, credited to usageFirst month waived. Auto-renews.Annual commit, paid up front, fully creditedRate card continues as overage
Usage ratesThe four published laddersBands reset each billing cycleSignificant discounts on every meter
Price / actionfrom $0.0030to $0.0003 at 10M+Custom on commit
vCPU · memory · GPUfrom $0.0230 · $0.0032 · $0.068Custom on commit
Monthly action ceilingNoneNone
Single-tenant control plane—$50k / yr
Guided Install (one-time)Standard $5,000 · Advanced from $15,000Standard $5,000 · Advanced from $15,000
Deployment
Self-managed: Union control plane, you run your clustersNo feeNo fee
Union runs your clusters (optional)$500 / cluster / mo$500 / cluster / mo, capped at $2,500
Self-hosted—By contract
CloudsAll regions, hyperscalers and neocloudsAny cloud
On-prem and air-gapped—
Marketplace commit drawdown—
Limits
Active clusters2Unlimited
Multi-region and multi-cloud on one control plane—
Concurrent actions10,000Custom
Fanout (actions per run)200,000Custom
AppsCustom
Named artifacts and tracked runsUnlimitedEach version and each run billed as an actionUnlimitedEach version and each run billed as an action
Retention30 days90-day trendsCustom, contractual
Extended retention or deletion terms beyond 30 days—
Identity and security
LoginFederated login (Google and Microsoft)Federated login, SSO, SAML
SCIM / group sync
Audit-log export
HIPAA BAAOn requestOn request
Zero Trust architecture
Managed secrets
Customer-owned secrets backend (Vault and similar)—
CMEK (customer-managed encryption keys)—
BYOK (bring your own key)—
Networking
Public endpoints
BYO-VPC
Private control-plane endpoint—
Proxy-only egress—
Private image registry—
Platform
Pure-Python authoring
Reusable containers
Fast cold starts
Caching
Replay
Recover
Remote debugging
Remote image builds
Ray / Spark / Dask plugins
Triggers
Queues
User management
Volumes
Custom cloud configuration—
Custom RBAC and namespace topology—
Support
Support planStandard, 12×5Sev 1 in 4 business hoursPremier included, 24×7 for Sev 1–2
Mission Critical support—Paid upgrade: Advanced or PrioritySee union.ai/support
Control-plane availability99.9% SLA99.9% SLA
Forward-deployed engineers—
FAQ

Frequently asked questions

Every self-service signup is a Team account. In your first billing cycle the $950 minimum is waived as a usage credit, so you only pay for usage beyond $950. If you ask Union to run your clusters, the $500 fee is also waived on your first cluster that cycle; self-managed accounts never pay it. There is no card at signup. Add a payment method before the credit runs out or the month ends, and any unused credit carries into your paid account. If no payment method is on file by then, orchestration pauses until one is added. Your infrastructure and data are untouched and nothing is charged automatically.

It is pre-paid usage, not a fee on top. On the first day of each billing cycle your account is charged $950, plus the fee for any clusters Union runs for you, and issued $950 of usage credit. Usage draws the credit down. If you use less, the remainder expires at the end of the cycle. If you use more, the extra is billed at the rate card. You pay the greater of your usage and $950.

An action is one durable unit of work: an invocation of flyte.task, flyte.trace, or flyte.new_condition running end to end to completion, a cache hit, an artifact version, or a tracked run. Each counts as one action.

Each meter has a published ladder. The first units you use in a billing cycle are priced at the first band, the next at the second, and so on. The lower rate applies only to the units above the threshold, so crossing a band never makes the units below it more expensive. Bands reset at the start of each billing cycle and nothing carries over.

No. An action is counted regardless of time, whether it finishes in two seconds or runs for a month. Compute is metered separately, to the second, on the resources the action allocates.

No. Actions are counted only once the task, trace, or condition succeeds, whether that happens on the first try or after five retries.

No. Union’s built-in durability layer is inherent to all actions and doesn’t count toward action usage. Whether Union checks on a task once or a hundred thousand times, that’s still one action.

We report the allocated resources (CPU, memory, and GPU accelerator) from each container running the actions within your workflows and apply usage-based pricing down to the second.

Every accelerator sits in one of three classes, set by the median on-demand price of that GPU across the clouds that sell it, and each class has its own ladder. The class rate is the same whichever cloud you run on. Reserved or owned GPUs are priced as a flat fee per GPU per month with unlimited hours. Talk to sales for a quote.

Union hosts the control plane on every plan. If you also want Union to provision, upgrade and operate the clusters in your cloud account, it is $500 per cluster per month, capped at $2,500 however many clusters you run, and it is optional. If you would rather run the clusters yourself, there is no fee. New accounts get their first Union-managed cluster free in month one.

Yes. Union is built by the creators of Flyte and runs the same SDK. Point your existing code at your Union tenant and it runs with the Union control plane, Zero Trust security and observability, with no rewrite.

When you’re ready to scale past one team and the two active clusters Team supports, or when security and procurement ask for more: a private endpoint or customer-managed keys, custom RBAC across business units, retention beyond 30 days, or 24×7 support. Enterprise is priced on an annual commit that is fully credited to usage, with significant discounts on every meter.

Fanout is the total number of actions a run creates, while concurrency is how many of those actions are running at the same time.

Yes we do! Talk to our Startup Team to learn more.

Yes we do! Talk to our Public Sector Team to learn more.

Yes, this is available as part of a Union Enterprise plan.

Yes, this is one of the deployment options available as part of a Union Enterprise plan. Self-hosted deployment supports on-prem, hybrid, and air-gapped configurations.

Zero Trust means your data never transits Union.ai’s infrastructure. Workflow inputs, outputs, logs, and code bundles are served directly from inside your own cloud through a secure tunnel.

No. You deploy the Union operator into a Kubernetes cluster you manage, which securely communicates with the Union control plane to poll for work.

Yes. Union runs in your own AWS, GCP, Azure, or neo-cloud account, with your data staying inside it.

Need a sanity check? Walk us through what you’re building.