
Your AI gateway should add 0.5 ms, not 20
Antirion sits between your apps and every LLM provider, and it is built to never be the slow part. One Rust binary with the console built in: point any OpenAI or Anthropic SDK at it and get fallbacks across providers, budgets per team and key, PII and secrets redaction, caching, and a log of every request with its route and cost. On 4 vCPUs it adds 0.5 ms at 5,000 RPS and holds 14,000 RPS, 4x Bifrost. Open source and self-hosted.
Antirion is an open-source developer tool that acts as an intermediary between applications and LLM providers, optimizing performance to add only 0.5 ms latency at 5,000 requests per second. It features built-in caching, request logging, and support for multiple SDKs, all while ensuring data privacy through PII and secrets redaction.
A source that found nothing is a measurement. A source that has not run is a gap. Neither means the launch lacks the thing.