Pricing
Pricing for Lake and Metis
Orama's cloud, Lake, is where you design, train, and ship AI on your own data. Run it in a shared region and pay only for what you use, or have a private warehouse, a dedicated cluster, or an on-prem installation for a monthly fee that includes usage credit. Metis is priced for each enterprise, with our sales team.
Lake
Ways to run Lake
Every way to run Lake, as lake.live sells it. Backups are included in every offer, and usage beyond the included credit is billed at the published price list below.
Private warehouse
$1,000 a month
Includes $400 of usage credit each month (40% of the fee).
A dedicated warehouse on shared hardware, in a shared region.
- Your own warehouse, isolated from other customers
- Backups included
- Your own domain, white-labelled (available on request)
Ordered on lake.live.
Dedicated cluster
$5,000 a month
Includes $2,000 of usage credit each month (40% of the fee).
A dedicated Kubernetes cluster on the cloud provider and in the region you choose, including regions beyond the shared ones.
- Your own Kubernetes cluster
- Choose the cloud provider and the region
- Backups included
- Your own domain, white-labelled (available on request)
Ordered on lake.live.
On-prem
From $10,000 a month
Includes at least $4,000 of usage credit each month (40% of the fee).
Source available, with full service (Metis). On any cloud provider or on bare metal, with any accelerators, in any region.
- Any cloud provider or bare metal, with any accelerators
- Any region, and multi-region deployments with geo-based load balancing and failover
- Full service (Metis)
- Backups included
- Your own domain, white-labelled (available on request)
Ordered with the order form on lake.live. We plan it with you, and nothing is charged by sending it.
Prices are in US dollars. Card payment on lake.live is not open yet: send the order form and we will be in touch, and nothing is set up or charged until we have spoken. For a cloud provider, hardware, or region that is not listed, or a larger order, use the order form.
Lake
The published price list
In a shared region there is no plan to choose and no monthly fee: each kind of usage has one rate, and your bill is what you used at that rate. The other offers include usage credit each month, and usage beyond it is billed at these rates.
| Usage and what it counts | Rate |
|---|---|
| CPUThe vCPUs each of your workloads is scheduled with: engines, services, apps, builds, and runs. | $0.0015per vCPU-minute |
| RAMThe memory each of your workloads is scheduled with. | $0.00017per GiB-minute |
| GPU: L4Time a workload holds an L4 GPU. | $0.0108per GPU-minute |
| GPU: A10GTime a workload holds an A10G GPU. | $0.0215per GPU-minute |
| GPU: A100Time a workload holds an A100 GPU. | $0.03875per GPU-minute |
| GPU: H100Time a workload holds an H100 GPU. | $0.07875per GPU-minute |
| GPU: other classesAny other GPU class, billed at the A100 rate. | $0.03875per GPU-minute |
| GPU memoryThe memory of the GPUs your workloads hold. | $0.00025per GiB-minute |
| StorageRepositories, container images, datasets, transcripts, and drive files, averaged over the month. | $0.04per GiB-month |
| TransferData moved into and out of your repositories, in and out counted together. | $0.05per GiB |
| Speech to textAudio turned into transcripts, by the length of the audio. | $0.0072per minute |
| API and MCP callsEvery API or MCP call made to a workspace. | $2.00per million calls |
| PredictionsEach prediction answered by a model in your workspace. | $0.01per thousand predictions |
| Model tokens: GPT-5.5 inputTokens sent to GPT-5.5 through Lake's model gateway. | $6.50per million tokens |
| Model tokens: GPT-5.5 outputTokens GPT-5.5 returns through Lake's model gateway. | $39.00per million tokens |
| Model tokens: GPT-5 inputTokens sent to GPT-5 through Lake's model gateway. | $1.625per million tokens |
| Model tokens: GPT-5 outputTokens GPT-5 returns through Lake's model gateway. | $13.00per million tokens |
| Model tokens: GPT-4o inputTokens sent to GPT-4o through Lake's model gateway. | $3.25per million tokens |
| Model tokens: GPT-4o outputTokens GPT-4o returns through Lake's model gateway. | $13.00per million tokens |
| Model tokens: text-embedding-3-small inputText sent to the text-embedding-3-small embedding model through Lake's model gateway. | $0.026per million tokens |
| Model tokens: text-embedding-3-small outputOutput tokens of the embedding model; an embedding normally reports none. | $0.026per million tokens |
| VolumesDisk attached to your services and stores, by size and time. | $0.20per GiB-month |
Rates are in US dollars. Model tokens are priced per model, input and output apart.
Counted across your workspaces
Usage is added up across all of your account's workspaces. Storage is the average of the month's daily readings.
Measured by the second, billed by the minute
Compute and speech are measured by the second. A part minute is counted once, on the month's total, never per job.
Compute is what a workload is scheduled with
CPU, RAM, GPU, and GPU memory are billed for what each workload is given while it runs, whether it uses all of it or not.
One invoice a month
Each calendar month you get one invoice, with a line for each kind of usage. You pay it by card or bank on a payment page run by Stripe.
Three example months
What a 31-day month in a shared region costs at the rates above, for three sizes of usage, before any model tokens, which are priced per model. Your bill is what you use; these are examples, not quotes.
Small
A workspace engine of 1 vCPU and 2 GiB and a database of 0.5 vCPU and 1 GiB, all month; 30 build hours; 40 GiB stored, 120 GiB of transfer, 200 minutes of speech, 3 million calls, 200,000 predictions, and 30 GiB-months of volumes.
- CPU
- $101.79
- RAM
- $23.07
- Storage
- $1.60
- Transfer
- $6.00
- Speech to text
- $1.44
- API and MCP calls
- $6.00
- Predictions
- $2.00
- Volumes
- $6.00
- Total
- $147.90
Medium
Workloads of 6 vCPU and 16 GiB all month; 120 build hours; an app on one L4 GPU (24 GiB) with 4 vCPU and 16 GiB for 100 hours; 300 GiB stored, 800 GiB of transfer, 1,500 minutes of speech, 25 million calls, 3 million predictions, and 250 GiB-months of volumes.
- CPU
- $443.16
- RAM
- $138.96
- GPU: L4
- $64.80
- GPU memory
- $36.00
- Storage
- $12.00
- Transfer
- $40.00
- Speech to text
- $10.80
- API and MCP calls
- $50.00
- Predictions
- $30.00
- Volumes
- $50.00
- Total
- $875.72
Large
Workloads of 24 vCPU and 96 GiB all month; 600 build hours; a training job on one H100 GPU (80 GiB) with 16 vCPU and 128 GiB for 200 hours; 3 TiB stored, 4 TiB of transfer, 10,000 minutes of speech, 200 million calls, 50 million predictions, and 2,000 GiB-months of volumes.
- CPU
- $1,922.04
- RAM
- $995.76
- GPU: H100
- $945.00
- GPU memory
- $240.00
- Storage
- $122.88
- Transfer
- $204.80
- Speech to text
- $72.00
- API and MCP calls
- $400.00
- Predictions
- $500.00
- Volumes
- $400.00
- Total
- $5,802.48
Metis
Priced for each enterprise
Metis, the Enterprise Intelligence OS, comes as one prepackaged program. Our forward-deployed engineers run it with your decision owners, step by step, on infrastructure you choose: your cloud, on-prem, or dedicated. We know what comes next at every step, so your team does not have to design the engagement.
Software: Zetta
The platform Metis is built on: one platform for everything that runs on your data.
Service: forward-deployed engineers
Our people, working inside your business unit, running the method on your data and your decisions.
The steps of the program
- 01
Inventory the decisions
The decisions that matter most to the business unit, with their owners.
- 02
Score the data
Each decision's data, scored on the ten availability dimensions, with evidence.
- 03
Write the returns chain
The return, written down and agreed before the work starts, so you can hold us to it.
- 04
Raise availability
Connect what exists, and create the data that was never captured.
- 05
Run the instruments
Reports, models, answers, and agents, each one admitted when its data can carry it.
- 06
Measure and hand over
Every decision logged, outcomes measured, the score re-taken on a cadence, and the loop handed to your own teams.
How Metis is priced
Metis is priced for each enterprise, with our sales team. The first thing we do is measure your data against your decisions, and nobody can price that from a website.
What you can hold us to
The returns chain names the baseline, the metric, and the window before any work starts, and the program reports against it.
Questions
How Lake billing works
- What counts as usage?
- Eleven things: CPU, RAM, GPU, GPU memory, storage, transfer, speech to text, API and MCP calls, predictions, model tokens, and volumes. The table above says what each one counts.
- Is there a monthly fee?
- Not in a shared region: there are no plans and no tiers, and you pay for the usage you have, at the rates above. A private warehouse, a dedicated cluster, and on-prem have a monthly fee that includes usage credit worth 40% of the fee, and usage beyond the credit is billed at the same rates.
- How does billing work?
- You get one invoice a month from Lake, for the calendar month, with a line for each kind of usage. You pay it by card or bank on a payment page run by Stripe.
- Can Lake run on our own infrastructure?
- Yes. On-prem runs on any cloud provider or on bare metal, with any accelerators, in any region, with full service (Metis). Send the order form on lake.live and we plan it with you. A dedicated cluster runs on the cloud provider and in the region you choose.
- Are backups included?
- Yes, in every way to run Lake: shared, a private warehouse, a dedicated cluster, and on-prem.
- Can Lake run on our own domain?
- Yes, with a private warehouse, a dedicated cluster, or on-prem: Lake can run on your own domain, white-labelled. It is available on request, and we set it up with you. The shared regions run on lake.live.
- How is Metis priced?
- Metis is priced for each enterprise. It is a program we run with you, so we scope it with you on a call. Get in touch.
Start with a conversation
Tell us what you want to design, train, and ship, or which decisions matter most to your business. We will tell you what Lake costs for your usage, or what enrolling in Metis looks like for you.