AI platforms and infrastructure · Global AI Product Index

How to choose

Who this is for. Teams that need a place to run, route, or host models rather than a single end-user chat app, including platform engineers, ML and data teams, and buyers comparing hosted inference with self-hosted stacks.

Who this is not for. Individuals who only want a consumer assistant with no API or deployment work, and teams with no tolerance for GPU capacity planning, rate limits, or vendor terms.

Separate inference APIs, model marketplaces, local runtimes, and full cloud platforms; they solve different jobs and rarely substitute for one another.
Decide whether models must stay on your own machines, may run on a vendor's GPUs, or can cross a regional border, since this rules out whole groups.
For product work, check rate limits, supported regions, uptime terms, and whether you can pin a specific model version instead of a moving alias.
Local stacks shift cost to hardware and operations; hosted stacks shift it to usage bills and vendor terms, so compare total cost over time.

Model marketplaces and gateways

Catalogs and gateways that expose many models behind one API or UI. Unlike the inference clouds next, they mostly broker access to third-party models rather than host their own hardware.

AI products
Name	Company	Category	Business model	Launched	Updated
OpenRouter Running models yourself	OpenRouter	Platforms & Infrastructure	Paid	2023	2026-10-09
Hugging Face Running models yourself	Hugging Face, Inc.	Platforms & Infrastructure	Freemium	2016	2026-09-10
Replicate Running models yourself	Replicate	Platforms & Infrastructure	Paid	2019	2026-09-10
Poe Running models yourself	Quora	Platforms & Infrastructure	Freemium	2023-02	2026-07-30
Civitai Running models yourself	Civitai	Platforms & Infrastructure	Freemium	2022-11	2026-09-10

Fast inference clouds

Hosts optimized for low-latency or high-throughput generation. Unlike the gateways above, they run the hardware themselves and compete on speed; unlike the local runtimes below, you reach them over the network instead of on your own machine.

AI products
Name	Company	Category	Business model	Launched	Updated
Groq Running models yourself	Groq	Platforms & Infrastructure	Freemium	2024-02	2026-09-10
Fireworks AI Running models yourself	Fireworks AI	Platforms & Infrastructure	Paid	2023	2026-09-10
Together AI Running models yourself	Together AI	Platforms & Infrastructure	Paid	2023	2026-10-09
Cerebras Running models yourself	Cerebras Systems	Platforms & Infrastructure	Freemium	2024-08	2026-10-09
Baseten Running models yourself	Baseten	Platforms & Infrastructure	Paid	2019	2026-10-09
Fal.ai Running models yourself	fal	Platforms & Infrastructure	Paid	2023	2026-10-09
Modal Running models yourself	Modal Labs	Platforms & Infrastructure	Freemium	2022	2026-09-10

Local and self-hosted runtimes

Software you run on a laptop or your own GPUs, including local UIs. Unlike the hosted clouds above, nothing leaves your machines and you carry the hardware cost; unlike the app builders below, they serve models directly rather than composing them into applications.

AI products
Name	Company	Category	Business model	Launched	Updated
Ollama Running models yourself	Ollama	Platforms & Infrastructure	Open (open weights/source)	2023-07	2026-10-10
LM Studio Running models yourself	LM Studio	Platforms & Infrastructure	Freemium	2023	2026-11-02
Jan Running models yourself Questions over your own documents	Menlo Research	Platforms & Infrastructure	Open (open weights/source)	2023	2026-10-13
vLLM Running models yourself	vLLM Project	Platforms & Infrastructure	Open (open weights/source)	2023-06	2026-10-10
Open WebUI Running models yourself	Open WebUI	Platforms & Infrastructure	Open (open weights/source)	2023	2026-10-10

LLM app and agent builders

Builders for composing LLM apps and agent workflows on top of models. Unlike the runtimes above, they add orchestration and user-facing logic; unlike the enterprise clouds below, they are usually self-serve and independent of one hyperscaler.

AI products
Select for comparison	Name	Company	Category	Business model	Launched	Updated	My shortlist
	Dify Running models yourself Questions over your own documents	LangGenius	Platforms & Infrastructure	Open (open weights/source)	2023	2026-10-10

Enterprise cloud AI services

Hyperscaler services that wrap models with IAM, networking, and enterprise contracts. Unlike the builders above, they are tied to one cloud's identity, billing, and governance; unlike the sovereign vendors below, they are global US hyperscalers with a broad service catalog.

AI products
Name	Company	Category	Business model	Launched	Updated
Amazon Bedrock Running models yourself	Amazon Web Services	Platforms & Infrastructure	Paid	2023-09	2026-10-09
Amazon Nova Running models yourself	Amazon Web Services	Platforms & Infrastructure	Paid	2024-12	2026-09-04
Microsoft Foundry Running models yourself	Microsoft	Platforms & Infrastructure	Paid	2024-11	2026-10-09
Gemini Enterprise Agent Platform Running models yourself	Google	Platforms & Infrastructure	Paid	2021-05	2026-09-10

Sovereign and regional AI platforms

Vendors outside the US hyperscalers that sell their own model platforms, often with data-residency or sovereignty positioning. Unlike the enterprise clouds above, they are independent providers rather than one global hyperscaler, and they usually emphasise regional hosting and local contracts.

AI products
Name	Company	Category	Business model	Launched	Updated
Aleph Alpha Running models yourself	Aleph Alpha	Platforms & Infrastructure	Paid	2019	2026-09-10
Upstage Running models yourself Questions over your own documents	Upstage	Platforms & Infrastructure	Freemium	2020	2026-10-09
Nebius Running models yourself	Nebius Group N.V.	Platforms & Infrastructure	Paid	2024	2026-10-09
Scaleway Running models yourself	Scaleway SAS	Platforms & Infrastructure	Paid	2024	2026-10-09
Nscale Running models yourself	Nscale Limited	Platforms & Infrastructure	Paid	2025	2026-09-10
Gcore Running models yourself	G-Core Labs S.A.	Platforms & Infrastructure	Freemium	2025	2026-10-09
OVHcloud AI Endpoints Running models yourself	OVH Groupe SA	Platforms & Infrastructure	Freemium	2025	2026-10-12
Infomaniak AI Services Running models yourself	Infomaniak Network SA	Platforms & Infrastructure	Freemium	2024	2026-10-12
STACKIT AI Model Serving Running models yourself	Schwarz Digits Cloud GmbH & Co. KG	Platforms & Infrastructure	Paid	2025	2026-10-12

Your next step

Every link above keeps the evidence attached: the index shows each value with its source and the day it was checked.

Frequently asked questions

How are products grouped?

Groups follow deployment or workflow, not popularity, and each product appears in one section of this guide.

Does this guide rank or benchmark platforms?

No. It groups notable indexed products by job to be done and links to sourced records. It does not score, rank, or run latency or throughput benchmarks.

Where do the facts come from?

From each product record in the index, which cites official and independent sources along with a last-reviewed date.

What are the limitations of these platforms?

Hosted services depend on vendor capacity, pricing changes, and regional availability, while self-hosted stacks depend on your own hardware and operations. Confirm current limits with the vendor before committing.

Related guides

Local and private document Q&A

Related comparisons

Sources and record dates

Record dates describe index updates, not independent product tests. Check linked documentation for current features and pricing.