Baseten

AI inference platform offering token-priced model APIs and dedicated GPU deployments of customers' own models, billed per minute and able to scale to zero.

www.baseten.copricingstatustermsdocs

Model weights in, an endpoint out.

Classification

Categories: Serverless, GPU cloud 1
Operating system: Self-managed 3
Use cases: AI and LLM apps, API backends 1

Tech stack

Runtimes: Python, Docker 3 4
GPU: Managed inference, Serverless GPU, Model API5 6 7

Pricing

Pricing model: Metered usage8
Entry price: $150 to $500 a month 8
Starting price: from $0.63 an hour8
Typical ceiling: Over $1,500 a month8
Currencies: USD 8
Billed: Monthly9
Paid: In arrears, for usage already run up9
Free tier: Trial only9

39 of 59 fields not recorded yet

Founded, Headcount, Reach, Favorite, Ownership, Parent, Wikidata, Metal, Runs on, CDN from, Audience, Control panel, Software, Specialised in, Shell, Managed databases, Persistent storage, Backups, Exit, Money back, Renewal multiple, Minimum contract, Payment, DPA, Energy claim, Certifications, Green Web directory, Domain registration, DNS hosting, Email hosting, CDN, Test domain, Collaboration, Staging, Support, Hours, Tiering, Published SLA, Infrastructure as code

Baseten sells AI inference in two shapes. Model APIs are OpenAI- and Anthropic-compatible endpoints over a catalog of open models the company runs, priced per million tokens. Dedicated deployments run the customer’s own model, packaged with the Truss framework from Python code and a config file, or as a custom Docker container running an inference server such as vLLM or SGLang, on GPUs billed per minute. Deployments autoscale, and the scaling docs show how to set the minimum to zero so an idle model releases its replicas. Training jobs on managed GPUs sit beside inference, and checkpoints can be deployed directly.

The self-serve plan has no monthly fee and charges usage; higher plans and self-hosted deployments are sold through sales. The contracting entity is Baseten Labs, Inc., in San Francisco.

Worth knowing

New workspaces receive credits, and the billing docs say models are deactivated when those run out and no payment method is on file. Invoices are issued when usage passes a threshold or at the end of the calendar month, whichever comes first. Deployments can be pinned to a United States or a European Union region, and regional selection requires a verified organization.

Identity

Status: Active

Deployment

Deployment: Container image 3

Regions and law

Regions: United States 10
Headquarters: United States 2

Automation

MCP: Official12
CLI: Official11
API: Public11

Sources

2026-09-24: Categories, Use cases
docs.baseten.co/overview
2026-09-24: Headquarters
www.baseten.co/terms-and-conditions/
2026-09-24: Operating system, Runtimes, Deployment
docs.baseten.co/development/model/custom-server
2026-09-24: Runtimes
docs.baseten.co/development/model/model-class
2026-09-24: GPU
docs.baseten.co/inference/model-apis/overview
2026-09-24: GPU
docs.baseten.co/development/model/build-your-first-model
2026-09-24: GPU
docs.baseten.co/deployment/manage/scaling
2026-09-24: Pricing model, Entry price, Typical ceiling, Starting price, Currencies
www.baseten.co/pricing/
2026-09-24: Billed, Paid, Free tier
docs.baseten.co/organization/billing
2026-09-24: Regions
docs.baseten.co/deployment/regional-deployments
2026-09-24: API, CLI
docs.baseten.co/deployment/manage/overview
2026-09-24: MCP
docs.baseten.co/agent-setup

Are you Baseten? Link back to this record.

The badge means this provider has a record on FindHost. It is not a rating, a rank, or an endorsement. It cannot be bought. A record exists whether the badge is displayed or not. More about the badge.

HTML, with the badge

Inline, so it loads nothing and takes the colour of the text around it.

<a href="https://www.findhost.app/baseten/">
  <svg xmlns="http://www.w3.org/2000/svg" width="120" height="44" viewBox="0 0 120 44" role="img" fill="currentColor">
    <title>Listed on FindHost</title>
    <rect x="0.5" y="0.5" width="119" height="43" rx="3" fill="none" stroke="currentColor" />
    <text x="60" y="17" text-anchor="middle" font-family="ui-sans-serif, system-ui, sans-serif" font-size="8" letter-spacing="1.5">LISTED ON</text>
    <text x="60" y="35" text-anchor="middle" font-family="Charter, 'Bitstream Charter', 'Sitka Text', Cambria, Georgia, serif" font-size="18" font-weight="600">FindHost</text>
  </svg>
</a>

HTML, text only

For a footer with no room for artwork.

<a href="https://www.findhost.app/baseten/">Listed on FindHost</a>

Markdown

For a README or a documentation page.

[Listed on FindHost](https://www.findhost.app/baseten/)

Updated 2026-09-24 · Listed since 2026-09-24Edit on GitHub This page as markdown

Somehow similar providers

Ordered by how many of category, regions, software, runtimes, entry price and operating system they share with this record. A field left blank on either side does not count.

Cerebrium
Serverless GPU platform for real-time AI workloads, running Python apps and custom Docker containers that scale to zero and bill per second.
Alibaba Cloud
Alibaba Cloud provides infrastructure, platform and serverless services spanning 32 regions with support for Linux, Windows, containers and multiple runtimes.
Appwrite
Open-source backend platform sold as a hosted service, running user functions alongside managed databases, authentication, storage and messaging.
Beam
Serverless GPU platform where Python functions and container images run as autoscaling endpoints, task queues and sandboxes, with on-demand GPU machines beside them.
Bunny.net
Slovenian edge platform selling CDN, storage, video and DNS, with Edge Scripting and Magic Containers running the customer's own code across its network.
Cherry Servers
Lithuanian provider selling dedicated machines and virtual servers on its own hardware, billed by the hour, by fixed term or as spot capacity.
Civo
British Kubernetes-first cloud that charges for worker nodes only, with control planes included and data transfer unmetered.
E2E Networks
Indian cloud infrastructure provider listed on the National Stock Exchange, selling CPU and GPU compute by the minute with published hourly rates.
Google Cloud Run
Google Cloud Run runs any container image, scales it to zero when idle and bills per request and per resource-second.
Koyeb
French platform that deploys containers and repositories across global regions with scale-to-zero, per-second billing, GPUs and serverless Postgres.
Latitude.sh
Bare-metal and GPU infrastructure billed hourly with automated provisioning, an API, SDKs, a CLI and Terraform support.
Modal
Serverless compute platform for Python, where functions are decorated in code and run in containers on CPU or GPU with per-second billing.
Nebius
Amsterdam-based AI cloud selling GPU and CPU virtual machines, managed Kubernetes and Slurm, serverless GPU containers and a model API.
Northflank
British platform running containers, jobs and managed databases on its own cloud, in the customer's cloud account, or on a self-hosted Kubernetes cluster.
Sesterce
French GPU cloud renting GPU virtual machines, bare-metal servers and inference endpoints by the hour from a prepaid balance.
SiteHost
New Zealand host that owns its Auckland data centre and sells Cloud Containers, a product that runs prebuilt or custom Docker images.
Amazon Web Services
Amazon Web Services is the largest cloud infrastructure provider, spanning 123 Availability Zones in 39 geographic Regions.
Arsys
Spanish hosting provider offering shared hosting and VPS from ISO-certified datacenters in Logroño, part of the IONOS Group since 2013.
Aruba.it
An Italian provider that owns its data centers and sells registrar services, hosting, servers and a public cloud.
AWS Lambda
Amazon's function-as-a-service primitive, running a handler on demand in a short-lived isolated environment and billing per invocation.