Last Updated: September 15, 2026
This Service Level Agreement (this “SLA”) applies to the the Services known as Dedicated Inference and Model APIs, in each case to the extent deployed on infrastructure managed by Baseten (the "SLA Service"). This SLA does not apply to Services deployed on infrastructure that Customer controls, including self-hosted deployments. This SLA is incorporated into and forms part of the Baseten Terms and Conditions available at https://www.baseten.co/terms-and-conditions/, or other written agreement between Baseten Labs, Inc. ("Baseten") and Customer governing Customer's use of the Services (in either case, the "Agreement"). Capitalized terms used but not defined in this SLA have the meanings given in the Agreement.
1. Definitions
The following capitalized terms will have the definitions set forth below:
1.1 "Force Majeure" means any act, event, or occurrence beyond Baseten's reasonable control, including, without limitation, issues arising from bugs or other problems in the software, firmware or hardware of Baseten's suppliers, outages or issues with upstream providers or network carriers, acts of God, fires, floods, storms, landslides, epidemics, lightning, earthquakes, drought, blight, famine, quarantine, blockade, governmental acts or inaction, orders or injunctions, war, insurrection or civil strife, sabotage, explosions, labor strikes, work stoppages, and acts of terror.
1.2 "Scheduled Downtime" means the total amount of time during any calendar month, measured in minutes, during which Customer is not able to access the SLA Service due to planned system maintenance performed by Baseten. Baseten will provide Customer with reasonable prior notice of such Scheduled Downtime.
1.3 "Total Monthly Time" means the total minutes in the relevant calendar month less Scheduled Downtime. For any partial calendar month during which Customer subscribes to the SLA Service, availability will be calculated based on the entire calendar month, not just the portion for which Customer subscribed.
1.4 "Unscheduled Downtime" means the total amount of time during any calendar month, measured in minutes, during which the Customer is not able to access the features and functions of the SLA Service as a result of a cause within Baseten's control, other than Scheduled Downtime. For the avoidance of doubt, and without limiting the foregoing, Unscheduled Downtime shall not include any period during which the SLA Service is unavailable as a result of (i) non-compliance by Customer with any provision of this SLA; (ii) incompatibility of Customer's equipment or software with the SLA Service; (iii) Customer's use of the SLA Service after Baseten has advised Customer to modify its use of the SLA Service, if Customer did not modify its use as advised; (iv) acts or omissions of Customer or Customer's employees, agents, contractors, or vendors, or anyone gaining access to the SLA Service by means of Customer's passwords or equipment; (v) performance of Customer's systems; (vi) any systemic Internet failures; (vii) network unavailability or Customer's bandwidth limitations; (viii) failures of any Third-Party Products, including any third party APIs; (ix) a Force Majeure; (x) Scheduled Downtime; or (xi) any error, failed or rejected request, throttling, rate-limiting, queuing, or reduced throughput resulting from (A) Customer failure to reserve an adequate number of GPUs, or sufficient committed capacity or throughput, for any SLA Service (including Customer’s election to purchase on-demand capacity instead of reserved capacity), or (B) Customer scaling above the number of any GPUs it reserved or any committed capacity or throughput for any SLA Service. For clarity, System Availability measures whether an SLA Service is reachable and operational; it does not measure Customer's ability to scale beyond reserved capacity or throughput.
1.5 "System Availability" means, with respect to any particular calendar month, the ratio obtained by subtracting Unscheduled Downtime during such month from the Total Monthly Time, and thereafter dividing the difference so obtained by the Total Monthly Time. Represented algebraically, System Availability for any particular calendar month is determined as follows:
System Availability = (Total Monthly Time - Unscheduled Downtime) / (Total Monthly Time)
2. System performance
Baseten will undertake commercially reasonable measures to ensure that System Availability of the SLA Service equals or exceeds ninety-nine point nine percent (99.9%) during each calendar month (the "Service Standard"). All measurements of System Availability will be calculated on a monthly basis for each calendar month during the Term based on Baseten's monitoring records, as published at https://status.baseten.co.
3. Customer requirements
Customer is responsible for maintenance and management of its computer network(s), servers, and software, and any equipment or services related to maintenance and management of the foregoing. Customer is responsible for correctly configuring its systems in accordance with any instructions provided by Baseten, as may be necessary for provision of access to the features and functions of the SLA Service.
4. Remedy
4.1 Service Credits. In the event the SLA Service fails to meet the Service Standard in a given month, Customer will be entitled to a service credit as set forth in the table below ("Service Credit"). Service Credits will be made in the form of a monetary credit applied to future use of the SLA Service.
4.2 Maximum Service Credits. The maximum amount of Service Credits that Baseten will issue to Customer for Unscheduled Downtime in a single calendar month will not exceed forty percent (40%) of the amount due by Customer for the applicable SLA Service for the applicable month.
4.3 Customer Must Request Service Credits. Customer must request such Service Credits by sending an e-mail identifying the date and time of the Unscheduled Downtime for which Customer is requesting Service Credits, with sufficient evidence (including description of the incident and duration of the incident) to billing@baseten.co within twenty four hours following such Unscheduled Downtime. If Customer fails to request any Service Credits to which Customer is entitled in accordance with this Section 4.3, Baseten will have no obligation to issue such Service Credits to Customer.