Baseten 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
Baseten suits teams running model inference who want a broad NVIDIA GPU menu and simple metering, with a GPU hour from $0.63 and per minute billing. It holds SOC 2 Type II and HIPAA and runs in 2 regions, so plans that need data residency beyond that come with a trade-off.
Key facts
- Best for
- Per-minute GPU inference on a focused NVIDIA menu, with SOC 2 Type II and HIPAA on file.
- GPU models
- NVIDIA T4, NVIDIA L4, NVIDIA A10G, NVIDIA A100, NVIDIA H100, NVIDIA H100 (MIG/fractional), NVIDIA H200, NVIDIA B200, NVIDIA RTX PRO 6000 (Source, checked September 26, 2026)
- GPU hour from
- $0.63 (Source, checked September 26, 2026)
- Billing granularity
- Per minute (Source, checked September 26, 2026)
- Locations
- 2 (Source, checked September 26, 2026)
- Interconnect
- High-speed InfiniBand between nodes in multi-node deployments (Source, checked September 26, 2026)
- Compliance certifications
- SOC 2 Type II, HIPAA (Source, checked September 26, 2026)
- Legal entity
- Baseten Labs, Inc. (Source, checked September 26, 2026)
- Headquarters
- United States (Source, checked September 26, 2026)
- Founded
- 2019 (Source, checked September 26, 2026)
- Support channels
- Email, Chat, Help-centre, Community (Source, checked September 26, 2026)
- Official website
- baseten.co
Serverless
- Compliance certifications
- SOC 2 Type II, HIPAA (Source, checked September 26, 2026)
- Legal entity
- Baseten Labs, Inc. (Source, checked September 26, 2026)
- Headquarters
- United States (Source, checked September 26, 2026)
- Founded
- 2019 (Source, checked September 26, 2026)
- Support channels
- Email, Chat, Help-centre, Community (Source, checked September 26, 2026)
Apps and profiles
- Elsewhere
- LinkedInXgithub.comyoutube.com
Pros and cons
Pros
- GPU hours start at $0.63, right at the site median of €0.63.
- Charging is per minute, which suits inference jobs that come and go in bursts.
- The list runs from entry-level cards through NVIDIA B200, with fractional H100 MIG for partial-GPU workloads.
- Multi-node deployments connect nodes over high-speed InfiniBand, useful for distributed jobs.
Cons
- Regional reach is 2 locations, below the comparison median of 4.
- Support runs on email, chat and help centre only, with no phone line.
- The card menu is shorter than ComputePrices.com's H100 SXM, A100 SXM and H200.
Who it suits
- Teams serving models in production that already handle their own orchestration and need SOC 2 Type II and HIPAA.
- Engineers who want a short, explicit GPU list and per minute billing rather than hourly meters.
- Buyers who prefer a narrow vendor: Baseten Labs, Inc., founded in 2019, sells dedicated GPU capacity.
Who should look elsewhere
- Shoppers who want consumer and AMD parts, which Salad covers through NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti.
- Buyers who need the older cards in SberCloud's B300, H100 and A100.
- Workloads that must stay in a market outside the 2 listed regions.
What it does well
Baseten runs NVIDIA T4, NVIDIA L4 and NVIDIA A10G, a focused list that reaches current datacenter parts rather than every consumer card. Deployments can be regional, and the provider lists 2 locations, so pin workloads to the nearest one before shipping. Multi-node setups connect nodes over high-speed InfiniBand, which is the part that matters for distributed jobs.
Who runs it
Baseten is Baseten Labs, Inc., a company founded in 2019 and headquartered in the United States. Support runs on email, chat and help centre, a self-service mix with chat in place of a phone desk, so plan on documented answers and community threads rather than live escalation for infrastructure issues. The company sells dedicated GPU capacity, not general-purpose cloud hosting.
How Baseten compares
Baseten vs BytePlus9.0 vs 8.7
Baseten vs RunPod9.0 vs 9.5
Baseten vs Netlify9.0 vs 8.6
Baseten vs ScalaHosting9.0 vs 8.3
Baseten vs Hostinger9.0 vs 9.1
Baseten vs Proton Drive9.0 vs 8.3
Baseten vs AIS Cloud9.0 vs 8.1
Where Baseten is available
- Available
- Everywhere except Belarus, Cuba, Iran, North Korea, Russia, Syria, Ukraine
Countries from the brand (Source, checked September 26, 2026)
Frequently asked questions
What does a GPU hour cost on Baseten?+
GPU hours start at $0.63, which is where the site median of €0.63 sits, and Baseten places 43 of 95 on that figure. The floor is set by the entry-level cards, so a job on a bigger card costs more per hour.
How does Baseten meter GPU usage?+
Usage is billed per minute, which is coarser than per-second metering. On the comparison, 26.4% still covers a minority of providers while 49.1% covers the bulk, so per-minute is a middle choice for bursty inference work.
Which GPUs can I run on Baseten?+
The published list is NVIDIA T4, NVIDIA L4 and NVIDIA A10G, and it includes a fractional H100 option for workloads that do not need a whole card. Salad's NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti is a consumer-first catalogue by comparison, while a much broader card shortlist sits at ComputePrices.com.
Where can I deploy Baseten workloads?+
Regional deployments cover 2 locations, against a comparison median of 4 and a top quartile of 10. Teams that must keep inference in a named country should check that list early, since fewer locations means less room to choose.
Is Baseten suitable for regulated workloads?+
Baseten lists SOC 2 Type II and HIPAA, which covers the usual enterprise security and health-data questionnaires. Teams with a named regulator or licence requirement still need to confirm the paperwork with the provider before signing, especially across the 2 listed regions.
How do I reach Baseten support?+
Support runs on email, chat and help centre, so a help centre and community carry the load, with chat for quick questions and no listed phone line. Across the comparison, 90.2% is widespread and 40.1% is much less so, so expect written answers first.
Who runs Baseten?+
Baseten is Baseten Labs, Inc., headquartered in the United States and founded in 2019. It sells dedicated GPU capacity for model inference rather than general-purpose cloud hosting, bills per minute, and lists 2 regional deployment locations.
Sources
- baseten.co/pricing checked September 26, 2026
Backs: Compliance certifications, Support channels
“Baseten is SOC 2 Type II certified and HIPAA compliant.”
- docs.baseten.co/deployment/regional-deployments checked September 26, 2026
Backs: Locations
“| `us` | United States | | `eu` | European Union |”
- docs.baseten.co/organization/billing checked September 26, 2026
Backs: Billing granularity
“Baseten meters usage by the minute while a workload is running on a node. The instance's per-minute price applies to serving replicas, image-builder workloads, and training workloads.”
- docs.baseten.co/performance/instances checked September 26, 2026
Backs: GPU hour from, GPU models, Interconnect
“| `T4x4x16` | $0.01052 | 4 | 16 GiB | 1 NVIDIA T4 | 16 GiB |”
- baseten.co/about-us checked September 26, 2026
Backs: Founded
“We started Baseten in 2019 after seeing the same failure over and over.”
- baseten.co/terms-and-conditions checked September 26, 2026
Backs: Headquarters, Legal entity
“Notices to Baseten shall be sent to Baseten Labs, Inc., 560 Davis St., Suite 250, San Francisco, CA 94111”
Compiled from 6 source pages, checked September 26, 2026. We did not open an account.
Not confirmed yet: Egress fee.
Related by the facts
Baseten
Per-minute GPU inference on a focused NVIDIA menu, with SOC 2 Type II and HIPAA on file.