Fal.ai 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
Fal.ai suits developers who need current GPU silicon metered by the second rather than the hour, and who can live with support by email, a help centre and community. The listed GPU hour price is $1.10, with H100, H200 and B200 on the list and SOC 2 in place.
Key facts
- Best for
- Per-second metering across a current GPU model list
- GPU models
- H100, H200, B200, B300, RTX PRO 6000, A100, A6000↗
- GPU hour from
- $1.10↗
- Billing granularity
- Per second↗
- Interconnect
- InfiniBand within a sector for 8xH100 multi-node instances↗
- Compliance certifications
- SOC 2↗
- Legal entity
- fal - Features & Labels, Inc.↗
- Headquarters
- United States↗
- Founded
- 2021↗
- Support channels
- Email, Help-centre, Community↗
- Pricing model
- Hourly billing
- Official website
- fal.ai
Facts checked September 26, 2026, 6 sources
Pros and cons
Pros
- Billing granularity is per second, so short jobs do not pay for unused time.
- Multi-node training can use InfiniBand within a sector for 8xH100 multi-node instances to link GPUs inside a sector.
- SOC 2 covers the platform, which buyers check before sending live data.
Cons
- The listed GPU hour price is $1.10, ranking 54 of 95 brands on that measure.
- Support runs on email, help centre and community only, with no live chat or phone line listed.
- The H100, H200 and B200 list is shorter than H100 SXM, A100 SXM and H200 at ComputePrices.com, which also covers older parts.
Who it suits
- Teams running short, intermittent jobs that want the meter to stop at the second.
- Groups scaling training across linked H100 nodes, as InfiniBand within a sector for 8xH100 multi-node instances describes.
- Developers who want SOC 2 in place before sending production traffic.
Who should look elsewhere
- ComputePrices.com, if you want the widest card catalog, H100 SXM, A100 SXM and H200, rather than a focused H100, H200 and B200.
- SberCloud, if its spread of B300, H100 and A100 fits your workload better.
- Salad, if your jobs suit the RTX and selected AMD classes in NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti.
What it does well
The catalogue is drawn from H100, H200 and B200, covering current datacenter silicon and workstation parts, so a team can pick a part by workload rather than settle for whatever is available. Larger training runs lean on InfiniBand within a sector for 8xH100 multi-node instances, which links the cards inside a multi-node instance within a single sector, letting nodes exchange data directly for distributed jobs.
Who runs it
Fal.ai is operated by fal - Features & Labels, Inc., headquartered in the United States and founded in 2021. Support is handled over email, help centre and community, so a help centre and a community space sit alongside direct email. No phone line or live chat desk is listed, which leaves written replies as the only route when a workload is stuck mid-run.
How Fal.ai compares
Fal.ai vs RunPod8.9 vs 9.5
Fal.ai vs ScalaHosting8.9 vs 8.3
Fal.ai vs Contabo8.9 vs 9.2
Fal.ai vs Kamatera8.9 vs 9.5
Fal.ai vs Cloudways8.9 vs 9.2
Fal.ai vs PlanetHoster8.9 vs 8.7
Fal.ai vs Acronis8.9 vs 9.9
Fal.ai vs Aruba8.9 vs 9.3
Fal.ai vs Decodo8.9 vs 9.1
Fal.ai vs Hostinger8.9 vs 9.1
Fal.ai vs MarsProxies8.9 vs 9.0
Fal.ai vs hide.me8.9 vs 8.7
Fal.ai vs Proton Drive8.9 vs 8.3
Frequently asked questions
What does a GPU hour cost on Fal.ai?+
The listed GPU hour price is $1.10. Across the brands tracked here, the median entry GPU hour is €0.63, and Fal.ai sits at position 54 of 95 on that measure, so it is priced in the middle rather than at the floor.
How finely is usage billed?+
Billing granularity is per second, so a job that finishes early does not carry a full hour of cost. The billing model is hourly, which suits workloads that start and stop with demand rather than run continuously.
Which GPUs can I choose from?+
The listed range is H100, H200 and B200, covering current datacenter parts and workstation cards. Readers wanting more choice, including older and consumer cards, will find a wider catalog at ComputePrices.com with H100 SXM, A100 SXM and H200.
Can training scale across more than one node?+
Yes, within a sector. InfiniBand within a sector for 8xH100 multi-node instances is the documented arrangement for multi-node instances, so a job can spread across the linked cards in that instance. Workloads that need to reach beyond one sector would need to check how instances are placed before committing.
How do I reach support?+
Support runs on email, help centre and community: direct email, a help centre for written questions, and a community space. There is no phone line or live chat desk listed, so teams needing a spoken conversation will need another route among the alternatives.
Who operates Fal.ai and where is it based?+
The service is provided by fal - Features & Labels, Inc., based in the United States and founded in 2021. A named contracting party matters for teams that need invoices raised against a specific business rather than an informal marketplace account.
Is there any compliance certification?+
The listed certification is SOC 2, and it is the only attestation on record here. Buyers who need a wider set of certificates or region specific evidence should confirm scope with the provider before signing.
How does the hardware list compare with the alternatives?+
Fal.ai runs on H100, H200 and B200, a focused datacenter and workstation set. SberCloud lists B300, H100 and A100, which reaches back to older parts, while Salad runs NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti, built on RTX and selected AMD classes. The right list depends on the silicon your jobs need.
Sources
- fal.ai checked September 26, 2026
Backs: Compliance certifications, Support channels
“fal is SOC 2 compliant and ready for enterprise procurement processes”
- fal.ai/about checked September 26, 2026
Backs: Founded, Headquarters
“Founded in 2021 by Burkay Gur and Gorkem Yurtseven”
- fal.ai/docs/documentation/compute checked September 26, 2026
Backs: Interconnect
“Instances within a sector are connected over InfiniBand, providing ultra-low latency and high bandwidth”
- fal.ai/docs/documentation/serverless/pricing checked September 26, 2026
Backs: Billing granularity
“you are billed for the total time your runners are alive, measured per-second by machine type”
- fal.ai/legal/terms-of-service checked September 26, 2026
Backs: Legal entity
“generative artificial intelligence media platform provided by fal - Features & Labels, Inc.”
- fal.ai/pricing checked September 26, 2026
Backs: GPU hour from, GPU models, Pricing model
“RTX PRO 6000 96GB $2.99/h $1.10/h”
Compiled from 6 source pages, checked September 26, 2026. We did not open an account.
Not confirmed yet: Egress fee.