Request access

Fantasti Cloud Platform Private Preview

Request access.

We are admitting a first cohort of companies while the platform is in Private Preview. Tell us what you want to run. Requests from personal email addresses join the waitlist for the next phase.

We reply to a reviewed request, usually within one business day.

Start from a request

01Capacity

GPU

Choose every type that would work. The first listing with a list price drives the estimate.

GPU count
Capacity
Term
Start
Interconnect
Region

Any country or region your data must stay in.

What will you run? Framework, model size and how it scales.

02Contact

Use your company email address. Requests from personal email addresses join the waitlist for the next phase.

Optional. One email asks you to confirm the address before any announcement is sent.

§01 Access First cohort

How an account is opened.

Access opens in cohorts. Companies can request a place in the first, and places are limited.

  1. 01 Request

    You send a request through the form, from your company email address, and say what you want to run.

  2. 02 Review

    The owner reads every request for the first cohort. A request from a company domain is reviewed for fit. We reply to a reviewed request, usually within one business day.

  3. 03 Invitation

    While places remain, an approved request is invited by a sign-up link sent to that address. The sign-up link is tied to the email address of the request, is valid for 72 hours and works once.

    • Valid for 72 hours
    • Works once
    • Tied to the email address of the request

  4. 04 Waitlist

    Personal and free email addresses join the waitlist in this phase. When the cohort is full, new company requests join the waitlist too.

§02 Reply Illustrative

How a request is placed.

Once your account is open, a request for capacity is checked the way the Fantasti Orchestrator checks every workload: against each pool's free GPUs, its fabric and its region. The record shows the pool that holds it and what turned the others away.

One request, checked and placed. Illustrative.

The request was 512 × H100 SXM5 · InfiniBand · us-* · reserved. 4 pools were checked. pool-b in us-east, fabric b was rejected: 256 free < 512. pool-c in us-west, no InfiniBand was rejected: no InfiniBand. pool-d in eu-west, fabric d was rejected: region not allowed. The job was placed in us-east, fabric a: 512 H100 SXM5 GPUs on one InfiniBand fabric. The record names the pool that can hold the request, us-east, fabric a, and its price line reads: reserved, rate in your order form. Connections: Request to Placement check: request; Placement check to The pool that holds it: placed.

  • TicketRequest6 mo reserved
    as the ticket wrote it
    gpu
    H100 SXM5
    count
    512
    fabric
    infiniband
    regions
    us-*
    capacity
    reserved
  • RecordPlacement check4 pools
    fx.clusters.create(gpu="H100:8", nodes=64, capacity=fantasti.Reserved("rsv_7d2k"))placement: { fabric: infiniband, regions: ["us-*"] }
    pool-b
    us-east · fabric b 256 free < 512
    pool-c
    us-west no InfiniBand
    pool-d
    eu-west · fabric d region not allowed
    pool-a
    us-east · fabric a 640 free
    free in pool-a 640asked 512

    Checked in order: GPU type, fabric, region, capacity class,free GPUs on one fabric. The first check a pool fails is thereason shown.

  • PlacementThe pool that holds it (highlighted)
    GPUs
    512
    Nodes
    64
    Fabric
    1
    HGX H100 8-GPU baseboard, line drawingHGX H100 · 8 GPUs per node
    placed · us-east · fabric a · 512×H100 SXM5 · one fabric
    price
    reserved · rate in your order form
  • TicketRequest6 mo reserved
    as the ticket wrote it
    gpu
    H100 SXM5
    count
    512
    fabric
    infiniband
    regions
    us-*
    capacity
    reserved
  • RecordPlacement check4 pools
    fx.clusters.create(gpu="H100:8", nodes=64, capacity=fantasti.Reserved("rsv_7d2k"))placement: { fabric: infiniband, regions: ["us-*"] }
    pool-b
    us-east · fabric b 256 free < 512
    pool-c
    us-west no InfiniBand
    pool-d
    eu-west · fabric d region not allowed
    pool-a
    us-east · fabric a 640 free
    free in pool-a 640asked 512

    Checked in order: GPU type, fabric, region, capacity class,free GPUs on one fabric. The first check a pool fails is thereason shown.

  • PlacementThe pool that holds it (highlighted)
    GPUs
    512
    Nodes
    64
    Fabric
    1
    placed · us-east · fabric a · 512×H100 SXM5 · one fabric
    price
    reserved · rate in your order form
    HGX H100 8-GPU baseboard, line drawingHGX H100 · 8 GPUs per node
  • TicketRequest6 mo reserved
    as the ticket wrote it
    gpu
    H100 SXM5
    count
    512
    fabric
    infiniband
    regions
    us-*
    capacity
    reserved
  • RecordPlacement check4 pools
    placement: { fabric: infiniband, regions: ["us-*"] }
    pool-b
    us-east · 256 free < 512
    pool-c
    us-west · no InfiniBand
    pool-d
    eu-west · region not allowed
    pool-a
    us-east · fabric a · 640 free
    free in pool-a 640asked 512
  • PlacementThe pool that holds it (highlighted)
    GPUs
    512
    Nodes
    64
    Fabric
    1
    HGX H100 8-GPU baseboard, line drawingHGX H100 · 8 GPUs per node
    placed · us-east · fabric a512×H100 SXM5 · one fabric
    price
    reserved
    rate in your order form
  • The pool that holds it
  • Your request
How capacity is placed
§03 Other requests

Not about capacity?