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 settingsCookie policy

Prime Intellect 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. Partners with a live offer in your country come first, then every other brand by score.

Our verdict

Prime Intellect spreads GPU capacity across from 50 locations and bills at per minute, so short jobs on current parts cost less than on hour-billed clouds. It suits small teams and independent researchers who want H200 and B300 class hardware without a long cloud contract.

Key facts

Best for
Small teams renting current datacenter GPUs with per-minute billing
GPU models
H200, B300, B200, H100, GH200, RTX Pro 6000, A100, A40↗
GPU hour from
$0.47↗
Billing granularity
Per minute↗
Locations
from 50↗
Interconnect
InfiniBand networking across nodes; multi-node clusters offer 3.2 Tbps InfiniBand (premium) or 100 Gbps Ethernet (value)↗
Legal entity
Prime Intellect, Inc.↗
Headquarters
United States↗
Founded
2024↗
Support channels
Email, Chat, Help-centre, Community↗
Pricing model
Hourly billing
Official website
primeintellect.ai

Serverless

Price per GB-second
$0.00000347↗

Facts checked September 26, 2026, 5 sources

Apps and profiles

Elsewhere
WikidataLinkedInX

Pros and cons

Pros

  • Billing granularity of per minute keeps short runs from paying for a whole hour.
  • The catalogue reaches current flagships: H200, B300 and B200.
  • Capacity spans from 50 locations, well above the 4 median in this group.

Cons

  • Support runs on email, chat and help centre only, so teams needing a phone line for incidents should look elsewhere.
  • Catalogue breadth is modest against ComputePrices.com, which tracks a much longer parts list (H100 SXM, A100 SXM and H200).
Prime Intellect

Ready to try Prime Intellect?

Scored 9.2 / 10 by DeployCue.

See alternatives

Who it suits

  • Small teams and solo researchers training on H200, B200 or B300 hardware
  • Workloads that need a choice of locations and node interconnect rather than a fixed region
  • Experiment builders who want to test logic on sandboxes before renting a card

Who should look elsewhere

  • Shoppers who want consumer-grade cards: Salad's list is built on RTX consumer parts and selected AMD classes (NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti).
  • Buyers who need V100: SberCloud lists it (B300, H100 and A100).

What it does well

Capacity is spread across from 50 locations, and the catalogue covers H200, B300 and B200, from A40 and A100 through H100 and H200 to B200, B300, GH200 and RTX Pro 6000. Node interconnect uses InfiniBand across nodes, with premium multi-node clusters on InfiniBand and value clusters on Ethernet.

Who runs it

The legal entity is Prime Intellect, Inc. (Prime Intellect, Inc.), a US company (US) founded in 2024, so contracts are with a young company rather than an incumbent. Support runs on email, chat and help centre, which suits developers comfortable with documentation and community threads.

How Prime Intellect compares

Where Prime Intellect is available

Available
Everywhere except Belarus, China, Cuba, Iran, North Korea, Russia, Syria, Ukraine

Countries from the brand (Source, checked September 26, 2026)

Frequently asked questions

How much does a GPU hour cost on Prime Intellect?+

GPU hours start at $0.47, below the site median of €0.63. Premium parts such as H200 and B300 can cost more than that entry figure, so read the number for the card you plan to use. On this measure Prime Intellect ranks 27 of 95 brands compared.

Which GPUs can I rent?+

The catalogue lists H200, B300 and B200, mixing older value cards with current flagships. A40 and A100 sit alongside H100, H200, GH200, B200, B300 and RTX Pro 6000, so a pipeline can start on a cheaper card and graduate to a faster card.

How is GPU usage billed?+

Billing granularity is per minute, so a job that finishes early does not pay for a whole hour. The pricing model is hourly, so charges follow the machine time you use rather than a fixed monthly commitment. Short fine-tuning and evaluation runs therefore cost less than on providers billing only by the hour.

Where do the GPUs run?+

Capacity is spread across from 50 locations, well above the 4 median for this group. Distributed setups can place nodes nearer to their own users, and multi-node clusters are connected with InfiniBand across nodes. Prime Intellect, Inc. is a US company (US).

What do sandboxes cost?+

Sandboxes are charged at $0.0000035 per GB-second, aimed at short experiments rather than sustained training. Because sandboxes and GPU rentals are separate products, code can be tested cheaply before any card is committed. Costs scale with the memory and time a sandbox actually uses.

Who runs Prime Intellect?+

The legal entity is Prime Intellect, Inc., headquartered in the US (US) and founded in 2024. Contracts sit with a young company rather than an incumbent hyperscaler, and the catalogue (H200, B300 and B200) mixes current flagships with older value cards, which keeps entry pricing low.

How do I get support?+

Support runs on email, chat and help centre. Chat and email cover routine questions about nodes and billing, the help centre documents sandboxes and cluster setup, and a community channel carries peer answers. Teams that need a phone line during out-of-hours incidents may prefer a larger cloud with wider service coverage.

Sources

  • docs.primeintellect.ai/faq checked September 26, 2026

    Backs: Billing granularity, Interconnect, Pricing model, Support channels

    “Each Pod has an hourly cost depending on the provider and GPU type. Your Prime Intellect credits are deducted every minute while the Pod is active.”

  • docs.primeintellect.ai/sandboxes/overview checked September 26, 2026

    Backs: Price per GB-second

    “Sandboxes are billed while running. For our new launch, these rates are valid through December 22, 2026: CPU: $0.02 per vCPU per hour; Memory: $0.0125 per GiB per hour (converted from $0.0125 per GiB...”

  • en.wikipedia.org/wiki/Prime_Intellect checked September 26, 2026

    Backs: Founded

    “Prime Intellect was co-founded in January 2024 by Vincent Weisser, a German decentralized science (DeSci) entrepreneur, and Johannes Hagemann”

  • primeintellect.ai checked September 26, 2026

    Backs: GPU hour from, GPU models, Locations

    “H200 ... $0.47/HR”

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

    Backs: Headquarters, Legal entity

    “We are registered in Delaware, United States”

Compiled from 5 source pages, checked September 26, 2026. We did not open an account.

Not confirmed yet: Egress fee.

Prime Intellect

Prime Intellect

Small teams renting current datacenter GPUs with per-minute billing

9.2/10See alternatives