Pricing

Powerful compute orchestration on our infrastructure or yours

Cloud

Serverless GPUs and AI infrastructure for advanced computation, model training, fine tuning, and inference

Start with $20 credit

Enterprise

Centralize infra management for all your users; deploy and orchestrate workflows across clouds and on-premises

Get a Demo

Open Source

Join a fast growing community and run your first workflow in less than 2 minutes

Learn more

Cloud Pricing

Only pay for the compute time utilized (data storage, memory, and networking are all included). No minimum commitment or monthly access fees. No reservations required – resources are available on-demand.

GPU
Quantity
Price (per GPU/hr)
Price (per GPU/sec)
H100 80GB
1, 2, 4, 8
$5.85 / hr
$0.0016250 / sec
A100 80GB
1, 2, 4, 8
$4.59 / hr
$0.0012750 / sec
V100
1, 4, 8
$3.67 / hr
$0.0010194 / sec
L40
1, 2, 4, 8
$2.22 / hr
$0.0006167 / sec
A10G
1, 4, 8
$1.21 / hr
$0.0003361 / sec
RTX A6000
1, 2, 4, 8
$1.46 / hr
$0.0004056 / sec
RTX A5000
1, 2, 4, 8
$0.90 / hr
$0.0002500 / sec
RTX A4000
1, 2, 4, 8, 10
$0.70 / hr
$0.0001944 / sec
T4
1, 4, 8
$0.64 / hr
$0.0001778 / sec
vCPU
–
$0.15 / hr
$0.00004167 / sec
Start with $20 credit

FAQs

We calculate usage based on the active time of compute resources, with charges based on the specific instance type utilized. For instance, consider a resource request with the following configuration: num_cpus=4, num_gpus=2, gpu_type="A100", memory="12GB", time_limit="3 hours". If the computation completes in 1 hour, 31 minutes, and 42 seconds, the cost calculation proceeds as follows.

Total Time Used: The active computing time amounts to 1 hour, 31 minutes, and 42 seconds, which equates to 5502 seconds.

Price Calculation: The cost is determined by multiplying 2 (the number of A100 GPUs) by $4.59 (the hourly rate for each A100 GPU, for example), and then multiplying this by 5502 seconds (the total active time), divided by 3600 seconds (the total number of seconds in an hour).

In this example, despite a time limit of 3 hours being set, the user is billed only for the 5502 seconds of resource utilization, amounting to approximately $14.06. If it is non-gpu request (that is, gpu_type is not provided, then the num_cpu alone will be priced based on CPU pricing).

Yes. For high-volume usage, Covalent Cloud provides discounts under our enterprise plan. Contact us to learn more about these options at contact@agnostiq.ai.

Payments are automated and charged to the credit card on file. Billing cycles are standardized and align with the end of each calendar month. If you join partway through a month, your first bill will be prorated based on the time of joining.

Yes. We offer special pricing for educational and non-profit projects. Reach out to us to explore these opportunities at contact@agnostiq.ai.