Glossary
B
Bandwidth allowance
A bandwidth allowance is the data a server may send and receive, and it decides when traffic slows down or costs extra.
Billing granularity
Billing granularity is the step a provider bills by, which decides how short a compute run can be before you pay for it.
Billing models
How a provider charges for usage, from a fixed monthly fee to hourly or per-second metering.
Bring your own key
Bring your own key means you supply your own model provider credentials to a gateway or proxy instead of paying the vendor to hold …
C
Compliance certifications
The audits and schemes a provider holds, which show how it handles security, privacy and regulated customers.
Control plane price
A control plane price is the hourly fee a managed Kubernetes provider charges for the cluster management layer, on top of the cost of …
Countries available
Countries available is the number of countries a proxy vendor can route traffic from, and it bounds how well you can localise a request.
D
E
Egress fee
A separate charge for outbound data, on top of the price of the instance or plan that produced it.
Egress price per GB
The cost of moving data out of a provider's network, charged per gigabyte sent.
Entry plan
The entry plan is the cheapest virtual server a provider sells, and its specs set the floor for what you get.
Entry price
The entry price is the headline cost of a provider's cheapest plan, and it shows how far your budget stretches.
F
Fee model
The shape of a provider's charges: per use, on a subscription, against a prepaid balance, or free software you host.
Founded
The year a provider was established, given as a four-digit year.
Free control plane
Whether the management layer that runs your clusters is billed separately or comes at no charge.
Free credits
Free credits are a prepaid balance handed to new sign-ups, so you can try paid features without paying first.
Free egress
Free egress means data leaving the provider is not charged, which matters most for anything that serves downloads or model responses.
Free egress (GB)
The gigabytes of outbound traffic a provider includes each month before it starts charging.
Free requests
The number of API calls a provider absorbs each month before per-request billing starts.
Free tier
A free tier is a level of service you can use at no cost, and it decides whether you can try before paying.
Free trial
A stretch of use at no cost, so a reader can test a provider before paying.
G
H
I
Input price from
The rate a provider charges for the tokens you send a model, quoted for a standard block of input tokens.
Interconnect
An interconnect is the network fabric that links compute nodes so data moves between them at memory speed rather than through a switch.
IP pool size
An IP pool size is how many addresses a proxy vendor says it can draw from, and a larger pool spreads blocking risk but …
ISO 27001
ISO 27001 is an audited information security management standard that cloud buyers ask for when customer data is handled.
L
Largest context window
The biggest prompt an AI model accepts in one request, measured in tokens.
Legal entity
The registered company that stands behind a provider's service, usually taken from its imprint or company register.
Locations
How many separate sites or regions a provider operates, which sets how close its servers can sit to your users.
M
Maximum run time
The longest a single serverless job may run before the platform stops it.
Minimum storage period
A minimum storage period is the shortest time an object storage vendor keeps your data before you can delete it, and the charge is …
Models offered
The list of models a provider serves, which sets what you can switch between from one API.
N
O
OpenAI-compatible API
An API that follows the OpenAI request and response format, so existing client code works unchanged.
Operating systems
Which operating systems a provider installs on its servers, most often Linux, Windows or both.
Output price from
The rate a provider charges for the tokens a model generates, which usually costs more per token than reading input.
P
PCI DSS
The payment card industry security standard, named in a provider's set of compliance certifications.
Points of presence
A point of presence is a single location where a network caches and serves traffic, and the count across the network tells you how …
Price from
The lowest price for a provider's entry paid plan, billed monthly.
Price per GB
What a provider charges for each gigabyte of data moving out of its network.
Price per GB per month
Price per GB per month is the storage rate for data you keep, quoted monthly so buckets of different sizes stay comparable.
Price per GB-second
A price per GB-second is the unit cost a serverless function provider charges for each gigabyte of memory it holds for each second it …
Price per request batch
The rate a provider charges for a large batch of API calls, the usual way endpoints price volume.
Pricing model
A pricing model is the rule a provider charges by, and it decides whether you pay for usage or for a plan.
Providers routed
Providers routed is how many model providers a gateway can forward requests to, and it sets how much of your stack it can absorb …
Proxy types
Proxy types describe where a proxy server sits and whose connection it carries, and that choice shapes the price, speed and detection risk of …
R
S
S3-compatible storage
S3-compatible object storage lets existing tools, SDKs and migration scripts work against a provider's buckets without changes.
Server types
The infrastructure categories a provider sells, such as virtual private servers and bare metal machines.
SLA uptime
SLA uptime is the share of time a provider commits to keeping your service reachable, and it backs most downtime promises.
SOC 2
SOC 2 is an outside auditor's report on the controls a company uses to protect data, with the type and period stated on it.
Storage price
A storage price is the cost of keeping a gigabyte of data at rest, usually quoted per month and separate from the cost of …
Storage types
Storage types are the kinds of disk storage a provider sells, and the mix decides what you can run on it.
Support channels
The ways a provider's customers can reach its support team, from email and phone to self-service help centres and community forums.