Inference providers
84 inference providers serving at least one tracked model through OpenRouter. Uptime is measured across each provider's endpoints over the last 30 minutes; each row averages that provider's deployments, and the summary gives every provider equal weight, so neither is a per-model guarantee. Entry price is the cheapest positive language-token output price the provider lists. Every provider name opens a page listing each model it serves with that provider's own price, context window, quantization and uptime.
Snapshot as of 7 November 2026
84 inference providers with at least one live endpoint for a tracked model as of 7 November 2026, observed through OpenRouter. Each provider name links to a page listing every model it serves with that provider's own price, context window, quantization and uptime.
DeepInfra, with 64 tracked models across 87 endpoints. Model count is unique canonical models, so a product served from several deployments counts once.
The lowest positive language-token entry price is $0.018 per 1M output tokens on PrimeIntellect. Entry price is the cheapest model a provider lists, not a like-for-like comparison; the cheapest-LLMs ranking and the compare tool line up specific models across providers.
Amazon Bedrock shows the highest mean measured uptime, 100.0%, over the last 30 minutes at the time of the snapshot; 82 of 84 providers report a measurement. Uptime is per endpoint from OpenRouter and averaged with equal weight, so it is a live signal rather than an SLA.
Both come from OpenRouter's per-endpoint catalogue and telemetry, re-observed on every ModelCap refresh. Prices are USD per 1M tokens as listed by the provider; uptime is the provider's own endpoint availability over the last 30 minutes. Offers retained from a failed refresh are marked stale and excluded from every count on this page.