Alpha
All entriesContact
Inference provider

Baseten

HQ
United States
Serves
8 open model families
Pricing
mixed
API
OpenAI-compatible

You substantially own inference here. The binding Terms and DPA bar training on your content (including deployed model weights), the Security Practices commit Model APIs and Dedicated Inference to Zero Data Retention, and confidentiality is mutual, a stronger standard-terms posture than most peers. A committed 99.9% SLA with credits adds reliability. Ownership falls short of full for four reasons. ISO 27001, PCI DSS and the SOC reports are vendor claims behind an access gate. The ZDR wording is a posture with a customer-controlled-factors carve-out and no explicit default. Async Inference keeps inputs for up to 72 hours. A self-serve EU region exists only for dedicated model deployments, and Model APIs residency needs a support-enabled policy. For EU residency of serverless traffic, or certificates confirmed first, deploy dedicated in the eu region or request the Trust Center documents.

Do you really own it?
Substantial
none·limited·partial·substantial·full
Analytical input: AOI B · 76.8/100
The four ownership factors

Floor-weighted, not averaged. Nothing is weak and both use & modify and data control are strong, so the floor is high; transparency sits at moderate, which is what keeps it short of full.

1

Use and modify freelyCan you use it freely and leave without lock-in?

Strong

OpenAI/Anthropic-compatible APIs over a predominantly open-weight Model Library keep workloads portable. The Terms & Conditions confirm the customer retains all IP in Customer Content, including deployed model weights and outputs (Sec 6.1). The deployment ladder (shared, dedicated, single-tenant, self-hosted, hybrid) gives control over where and how the workload runs.

How this scores (AOI sub-dimensions)
Transparency & lock-in4/5how portable it is and how easily you can leaveAn OpenAI-compatible API (plus Responses and Anthropic Messages endpoints) over a predominantly open-weight catalogue keeps workloads portable.
Cost4/5how the pricing model compares and how predictable it isPublic per-token rates for Model APIs and per-GPU-minute rates across a wide range of GPU classes (T4 through B200) are published on Baseten's pricing page, with defined tiers (Basic/Pro/Enterprise) and published default rate limits.
2

TransparencyAre the binding terms published, legible and independently checkable?

Moderate

The Terms & Conditions, DPA, Security Practices, SLA and docs are legible, and the Trust Center renders in a browser (13-row sub-processor list, dated updates feed). Held at moderate because every certificate and report behind it (SOC 2 and SOC 3, ISO 27001, HIPAA, pen test, PCI AOC, compute-provider roster) is access-gated. The Security Practices have no explicit 'ZDR is the default' sentence and carry a customer-controlled-factors carve-out, and the docs do not say where the control plane runs.

How this scores
Not a scored AOI dimension. For a hosted provider, transparency is whether the binding terms are published, legible and were actually read - the read/unverified evidence below, not a certification. A strong rating here must trace to a retrieved binding document.
3

ReliabilityDoes it stay up and stay secure?

Strong

The SLA commits to 99.9% with service credits capped at 40% and a 24h claim window, and is worded as commercially reasonable measures. It is measured on a public status page that logs 7 major or critical incidents between 2026-09-20 and 2026-11-18, all resolved, including a 21-minute outage of all Model APIs on 2026-11-11. Reliability dimension 4 maps to strong.

How this scores (AOI sub-dimensions)
Reliability4/5whether it stays up, with an SLA and status historyA committed SLA (baseten.co/service-level-agreement, stamped September 15, 2026) undertakes "commercially reasonable measures" to reach 99.9% monthly availability for both Dedicated Inference and Model APIs.
Security4/5the controls protecting your traffic and dataIsolation is logical namespace separation for shared use (access controls, unique customer identifiers, namespace isolation), single-tenant dedicated Kubernetes namespaces with Calico/Cilium network policies ("GPUs never shared across users") and self-hosted deployment inside the customer's own VPC.
Compliance4/5which independent certifications and attestations it holdsSOC 2 Type II is confirmed with a named audit firm (Sensiba San Filippo LLP, clean opinion across Security/Availability/Processing Integrity/Confidentiality/Privacy criteria).
4

Doesn't extract your dataDo the binding terms keep your data and IP yours?

Strong

Terms Sec 6.3 and DPA 4.4 bar training, and the Security Practices state a ZDR posture for Model APIs and Dedicated Inference (docs: 'Synchronous requests run under ZDR by default'). Customer Content is customer-owned, and Sec 11 sets a mutual confidentiality duty (data_governance 4). Caveats: Async Inference queues inputs for up to 72 hours, the Security Practices exempt retention that follows from customer configuration or customer code, and Model APIs residency needs a support-enabled policy.

How this scores (AOI sub-dimensions)
Data governance4/5retention, training-on-inputs and data ownershipThe binding documents set a strong posture.
Residency3/5where your data is processed and storedAn EU region is available for at least some workloads, which is the score-3 anchor.

How the AOI score is computed

The seven dimensions above, each scored 0 to 5, weighted and summed to the 0 to 100 headline. The score is the analytical input behind the ownership verdict, not the verdict itself.

DimensionScoreWeightPoints
Data governance4/50.2419.2
Compliance4/50.1814.4
Residency3/50.169.6
Security4/50.1411.2
Reliability4/50.129.6
Transparency & lock-in4/50.108.0
Cost4/50.064.8
HeadlineB · 76.8/100
Dossier coverageAssess 81%Implement 63%Use 79%Support 57%How complete our four-domain documentation is, a measure of our coverage, not of the model. Each domain links to its page.

Sources

Every rating traces to a primary document. Read means the text was verified; unverified means it is known to exist but has not yet been read.

DocumentWhat it grounds
Terms of serviceread2026-11-19
Baseten Terms & Conditions.
Securityread2026-11-19
Baseten Security Practices (Last Updated November 12, 2026).
Data Processing Addendumread2026-11-19
Baseten DPA (Last Updated September 15, 2026).
Vendor announcementread2026-11-19
Baseten blog post (originally published 2023-03-08, last updated 2026-11-02; vendor marketing) stating SOC 2 Type II, audited by Sensiba San Filippo LLP, covering Security/Availability/Processing Integrity/Confidentiality/Privacy criteria with a clean opinion across 70+ sub-criteria.
Securityread2026-11-19
Baseten Trust Center (Vanta; launched 2026-07-30; renders only in a browser, curl returns only a title).
Documentationread2026-11-19
Baseten Regional deployments docs.
Documentationread2026-11-19
Baseten regional-environments docs and the Model APIs overview.
Documentationread2026-11-19
Baseten hosting options and Self-Hosted docs (hosting-options/overview.md and observability/security.md).
Documentationread2026-11-19
Baseten security docs (docs.baseten.co/security.md and observability/security.md).
Slaread2026-11-19
Baseten status page (status.baseten.co statuspage JSON): 'All Systems Operational' across Dedicated Inference, Model APIs, Training, Model Management API, Web Application, Homepage and Docs, and Sandboxes (with Oregon, Virginia and Frankfurt Sandboxes components).
Slaread2026-11-19
Baseten Service Level Agreement (Last Updated September 15, 2026).
Vendor announcementread2026-11-19
Baseten pricing page.
Documentationread2026-11-19
Baseten Model APIs overview (docs.baseten.co/inference/model-apis/overview.md).
Documentationread2026-11-19
Baseten Model APIs pricing-and-limits and deprecation docs.
Documentationread2026-11-19
Baseten platform overview docs (docs.baseten.co/overview.md and the Model APIs overview).