About AI Inference Providers
Every inference API, priced on the same model.
Choosing an inference API means reading a dozen pricing pages that each list different models in different units. AI Inference Providers does that reading for you: 28 live providers, from GPU clouds and custom-silicon chips to the hyperscalers and routers, priced on the same model (gpt-oss-120b) in the same unit (USD per 1M tokens), next to the facts a security review asks for.
Benchmark sites rank models; router pages list only the hosts they route to. This table is provider-first: serverless and dedicated pricing side by side, custom silicon against GPUs, OpenAI compatibility, SOC 2, HIPAA and zero data retention, and a clear flag on APIs that have shut down.
Every number links to the provider’s own page and every row shows the date we checked it. The methodology explains each field.
AI Inference Providers is published by Beundra AB, Sweden. Providers can buy one clearly labelled featured slot; the data and the sort order come from the sources alone. Questions and corrections: hello@aiinferenceproviders.com.