The independent guide to LLM inference providers

Serverless per-token APIs, dedicated endpoints, bring-your-own-model platforms, aggregators, and cloud model platforms. Find providers that match your models, data rules, and deployment model, with every fact linked to its source.

Explore the landscape Read the buyer’s guide

17vendors profiled: Managed services, expert networks, platforms, and dataset sellers.
90primary sources: Vendor sites, trust portals, and company releases. No listicles.
0paid placements: Nobody pays to appear or to change their position.

Who covers what

Filter vendors

Published workload coverage for 17 LLM inference providers. Click a column to see those vendors.

Which workloads each vendor publicly describes
Vendor	OpenAI-compatible	Open-weight models	Proprietary models	Custom models	Zero-retention options	Region control	Batch inference	Custom silicon
Amazon Bedrock	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No
Baseten	Yes	Yes	No	Yes	Yes	Yes	No	No
Cerebras	Yes	Yes	No	Yes	No	No	No	Yes
Cloudflare Workers AI	Yes	Yes	No	Yes	Yes	No	Yes	No
DeepInfra	Yes	Yes	Yes	Yes	Yes	No	No	No
Fireworks AI	Yes	Yes	No	Yes	Yes	No	Yes	No
Google Cloud Vertex AI (Gemini Enterprise Agent Platform)	Yes	Yes	Yes	Yes	Yes	Yes	No	No
Groq	Yes	Yes	No	Yes	Yes	No	Yes	Yes
Hugging Face Inference Providers	Yes	Yes	No	No	Yes	No	No	No
Hyperbolic	Yes	Yes	No	No	No	No	No	No
Microsoft Foundry (formerly Azure AI Foundry)	Yes	No	Yes	Yes	No	Yes	Yes	No
Modal	Yes	No	No	Yes	No	Yes	No	No
Nebius Token Factory (formerly Nebius AI Studio)	Yes	Yes	No	Yes	Yes	Yes	No	No
OpenRouter	Yes	Yes	Yes	No	Yes	Yes	Yes	No
Replicate	No	Yes	Yes	Yes	No	No	No	No
SambaNova	Yes	Yes	No	No	No	Yes	No	Yes
Together AI	Yes	Yes	No	Yes	No	No	Yes	No
Vendors	16	15	6	13	10	8	7	3

Based on each vendor’s published categories and capabilities. A filled dot means the vendor describes this workload publicly; it is not a quality judgment.

Vendor landscape

17 providers, checked November 8, 2026. Filter by workload, buying model, published certifications, and public pricing. Select up to four to compare side by side.

LLM inference providers in alphabetical order with buying model, ownership, published certifications, and pricing
Vendor	Buying model	Ownership	SOC 2	ISO 27001	HIPAA	GDPR	Pricing
Amazon BedrockCloud provider platform	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model, Cloud provider platform	Subsidiary	Not found	Not found	Published	Published	Published
BasetenServerless per-token API (Model APIs)	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model	Privately held	Published	Published	Published	Published	Published
CerebrasServerless per-token API	Serverless per-token API, Dedicated or provisioned capacity	Publicly listed	Published	Not found	Not found	Published	Published
Cloudflare Workers AIServerless per-token API on Cloudflare's network	Serverless per-token API	Publicly listed	Published	Published	Published	Published	Published
DeepInfraServerless per-token API	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model	Privately held	Published	Published	Published	Not found	Published
Fireworks AIServerless per-token API	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model	Privately held	Published	Published	Published	Published	Published
Google Cloud Vertex AI (Gemini Enterprise Agent Platform)Cloud provider platform	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model, Cloud provider platform	Subsidiary	Published	Published	Published	Not found	Published
GroqServerless per-token API	Serverless per-token API	Privately held	Published	Not found	Published	Published	Published
Hugging Face Inference ProvidersAggregator / router across inference providers	Serverless per-token API, Aggregator / router	Privately held	Published	Not found	Not found	Published	Published
HyperbolicServerless per-token API	Serverless per-token API	Privately held	Published	Not found	Not found	Not found	Published
Microsoft Foundry (formerly Azure AI Foundry)Cloud provider platform	Serverless per-token API, Dedicated or provisioned capacity, Cloud provider platform	Subsidiary	Not found	Not found	Not found	Not found	Published
ModalServerless GPU compute	Serverless per-token API, Bring your own model	Privately held	Published	Not found	Published	Published	Published
Nebius Token Factory (formerly Nebius AI Studio)Serverless per-token API	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model	Publicly listed	Published	Published	Published	Not found	On request
OpenRouterAggregator (routes to many providers)	Aggregator / router	Privately held	Published	Not found	Not found	Not found	Published
ReplicateServerless model API (per output, per token, or per second)	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model	Subsidiary	Not found	Not found	Not found	Not found	Published
SambaNovaServerless per-token API (SambaCloud)	Serverless per-token API, Dedicated or provisioned capacity	Privately held	Published	Published	Not found	Not found	Published
Together AIServerless per-token API	Serverless per-token API, Dedicated or provisioned capacity, Bring your own model	Privately held	Published	Published	Not found	Not found	Published

Alphabetical. A check means the vendor publishes it on its own site or trust portal; a dash means we found nothing public, so ask the vendor.

Research and comparisons

GuideNov 8, 2026 How to choose an LLM inference provider Seven decisions, in order: workload, deployment model, API compatibility, data handling, regions, security evidence, and commercials. ComparisonNov 8, 2026 Together AI vs Fireworks AI ComparisonNov 8, 2026 Groq vs Cerebras ComparisonNov 8, 2026 Amazon Bedrock vs Google Cloud Vertex AI

Browse by workload

How this guide works

Every claim links to a public source, with the date we checked it.
No paid rankings and no speed or quality scores. Providers are listed alphabetically or by the filters you choose.
Pricing: we say where a provider publishes it and quote one example. Prices change often, so check the source.

Read the methodology