Your cookie choices

We use necessary cookies to make this site work. With your permission, analytics cookies help us understand usage and marketing cookies support advertising.

You can reject non-essential cookies and still use the site. Change your choices anytime in Cookie settings.

Cookie policy
Cookie settings

Cerebrium review

Updated September 28, 2026 · DeployCue Editorial Team · Facts checked September 26, 2026

We earn a commission when you sign up through links on this site. Partner status can affect where a brand appears in our rankings.

Our verdict

Cerebrium suits developers and small teams who need serverless inference and fine-tuning without holding bare metal. Its free tier plan pairs $0.59 per GPU hour with per second billing, and the offer score of 8.4 sits well above the cost score of 5.1.

Key facts

Best for
Serverless GPU inference and fine-tuning on a meter that bills per second.
GPU models
NVIDIA B200, NVIDIA H200, NVIDIA H100, NVIDIA RTX PRO 6000, NVIDIA A100 80GB, NVIDIA A100 40GB, NVIDIA L40s, NVIDIA L4, NVIDIA A10, NVIDIA T4, AWS Inferentia2, AWS Trainium (Source, checked September 26, 2026)
GPU hour from
$0.59 (Source, checked September 26, 2026)
Billing granularity
Per second (Source, checked September 26, 2026)
Free credits
First 100GB storage free (Source, checked September 26, 2026)
Locations
12 (Source, checked September 26, 2026)
Compliance certifications
SOC 2, HIPAA, GDPR, ISO 27001 (Source, checked September 26, 2026)
Legal entity
Cerebrium Inc (Source, checked September 26, 2026)
Headquarters
United States (Source, checked September 26, 2026)
Founded
2021 (Source, checked September 26, 2026)
Support channels
Email, Chat, Community, Help-centre (Source, checked September 26, 2026)
Pricing model
Free tier
Official website
cerebrium.ai

Serverless

Price per GB-second
$0.00000222 (Source, checked September 26, 2026)
Compliance certifications
SOC 2, HIPAA, GDPR, ISO 27001 (Source, checked September 26, 2026)
Legal entity
Cerebrium Inc (Source, checked September 26, 2026)
Headquarters
United States (Source, checked September 26, 2026)
Founded
2021 (Source, checked September 26, 2026)
Support channels
Email, Chat, Community, Help-centre (Source, checked September 26, 2026)

Apps and profiles

Elsewhere
LinkedInXgithub.com

Pricing

  • BLACKWELL RTX6000 (USD)

    compute: $2.00 / hour

  • TRN1 (USD)

    compute: $1.80 / hour

  • INF2 (USD)

    compute: $1.01 / hour

  • Cerebrium NVIDIA T4 (USD)

    compute: $0.59 / hour

  • Cerebrium NVIDIA L40 (USD)

    compute: $1.95 / hour

  • Cerebrium NVIDIA L4 (USD)

    compute: $0.80 / hour

  • Cerebrium NVIDIA H200 (USD)

    compute: $3.30 / hour

  • Cerebrium NVIDIA H100 (USD)

    compute: $2.21 / hour

  • Cerebrium NVIDIA B200 (USD)

    compute: $6.25 / hour

  • Cerebrium NVIDIA A100 80GB (USD)

    compute: $2.50 / hour

  • Cerebrium NVIDIA A100 40GB (USD)

    compute: $2.10 / hour

  • Cerebrium NVIDIA A10 (USD)

    compute: $1.10 / hour

Pros and cons

Pros

  • Hardware choice is broad for a single vendor: NVIDIA B200, NVIDIA H200 and NVIDIA H100.
  • Billing is per second, so idle capacity is not charged as a whole hour.
  • Certifications are SOC 2, HIPAA and GDPR, and storage starts with a free allowance.
  • Deployments span 12 regions, above the GPU median of 4.

Cons

  • The entry rate of $0.59 per GPU hour is near the €0.63 median, so it gives no clear price edge over the field.
  • Support runs on email, chat and community only, with no phone channel offered.
  • Billing this finely appears on only 26.4% of tracked brands, so hourly-billed plans are easier to compare.
Cerebrium

Ready to try Cerebrium?

Scored 9.3 / 10 by DeployCue.

See alternatives

Who it suits

  • Developers shipping inference endpoints and fine-tuning jobs straight from code, without provisioning machines.
  • Buyers whose compliance checklist expects SOC 2, HIPAA and GDPR.
  • Small teams watching a meter, since billing is per second and storage starts free.

Who should look elsewhere

  • Readers after a B300 should look at SberCloud, whose B300, H100 and A100 list includes it.
  • Buyers who want the longest hardware catalogue should check ComputePrices.com with H100 SXM, A100 SXM and H200.
  • Shoppers after consumer and gaming cards should look at Salad, whose NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti is a different class from Cerebrium's datacenter parts.

How Cerebrium compares

Frequently asked questions

What does a GPU hour cost on Cerebrium?+

The free tier plan starts at $0.59 per GPU hour. The site median across tracked brands is €0.63, and Cerebrium ranks 41 of 95 on that figure, so the entry rate sits close to the middle of the field.

How finely is usage measured?+

Cerebrium bills per second, the finest granularity the site records. Only 26.4% of tracked brands charge that way, so short jobs and bursty inference traffic avoid paying for a whole hour.

Which GPUs can workloads run on?+

The catalogue is NVIDIA B200, NVIDIA H200 and NVIDIA H100. It pairs current datacenter silicon with older cards and AWS accelerators, giving both fine-tuning and inference workloads a hardware choice without a second provider. A B300, listed by SberCloud in B300, H100 and A100, is the gap to check first.

Is there a free way to try it?+

The entry plan is free tier, and storage carries an introductory free allowance, so a first deployment can be tried at no cost. The site median GPU hour price is €0.63, so a free start is not the only way in.

Where can a model be deployed?+

Deployments can be spread across 12 regions, above the GPU field median of 4 and under the upper quartile of 10. Cerebrium is headquartered in the US and contracts as Cerebrium Inc.

Which certifications does it hold?+

The stated certifications are SOC 2, HIPAA and GDPR, which covers the usual procurement checklist for a US vendor. Cerebrium was founded in 2021 and is listed as a US company, so contract paperwork runs through Cerebrium Inc.

How do you reach support?+

Support runs on email, chat and community, so help comes by email, chat, the community or the help centre, with no phone line among the listed channels. That suits developers who can describe an issue in a thread.

What does serverless-style compute cost?+

Compute time is priced per GB-second at $0.0000022, which suits short, uneven request patterns rather than steady loads. Storage begins with a free allowance, so the first part of a project's storage is not billed.

Sources

  • cerebrium.ai/about checked September 26, 2026

    Backs: Headquarters

    “Founded in Cape Town, South Africa and now headquartered in New York City”

  • cerebrium.ai/pricing checked September 26, 2026

    Backs: Billing granularity, Compliance certifications, Free credits, GPU hour from, Pricing model, Price per GB-second, Support channels

    “Cerebrium charges based on actual compute time, measured in seconds.”

  • cerebrium.ai/terms-of-service checked September 26, 2026

    Backs: Legal entity

    “These Terms of Service ... govern your use of our web pages located at https://www.cerebrium.ai operated by Cerebrium Inc.”

  • docs.cerebrium.ai/docs/deployments/multi-region-deployment checked September 26, 2026

    Backs: Locations

    “Contact support to enable access to these regions: us-west-2, eu-west-2, eu-central-1, ap-south-1, ap-northeast-1, sa-east-1, ca-central-1, me-central-1”

  • docs.cerebrium.ai/docs/hardware/using-gpus checked September 26, 2026

    Backs: GPU models

    “The platform offers GPUs ranging from cost-effective development options to high-end enterprise hardware, alongside CPU-only compute and AWS accelerators.”

  • ycombinator.com/companies/cerebrium checked September 26, 2026

    Backs: Founded

    “Founded:2021 Batch:Winter 2022 ... Location:New York City, NY”

Compiled from 6 source pages, checked September 26, 2026. We did not open an account.

Not confirmed yet: Egress fee.

Cerebrium

Cerebrium

Serverless GPU inference and fine-tuning on a meter that bills per second.

9.3/10See alternatives