Hyperion
The enterprise-grade AI gateway for production LLM apps.
Is this your product? Claim it to edit the listing, post updates and answer questions. Prove it with an email address at the product's domain (the listing becomes Claimed), or put the LotsDirectory badge on your website (the listing becomes Verified, and its link to your website is followed while the badge is live).
Description
The enterprise-grade AI gateway for production LLM apps. Unified API, microsecond-latency caching, prompt security, and real-time PII redaction.
About Hyperion
What is Hyperion?
Hyperion is an open source AI gateway written in Go that provides multi-provider model routing, caching, and rate limiting.
What Hyperion does
Hyperion routes requests across 190+ endpoints for providers including OpenAI, Anthropic, Google, and Azure through an OpenAI-compatible API. The gateway implements multi-layer caching with Redis exact matches and Qdrant semantic caching, automatic failover, spend forecasting, and PII redaction.
Who Hyperion is for
Built for developers, ML teams, and enterprise deployments running scale-ready AI and production LLM applications.
How Hyperion works
Developers deploy Hyperion as a self-hosted single binary or connect through cloud endpoints. The gateway accepts OpenAI-compatible API calls, checks cache layers, routes requests across model providers, and applies budget cutoffs.
Key features
- Unified OpenAI-compatible API abstracting endpoints across OpenAI, Anthropic, Google, and Azure.
- Multi-layer caching combining Redis exact matching and Qdrant semantic embedding search.
- Written in Go with sub-millisecond gateway execution overhead.
- Automatic failover and circuit breaking across model providers without changing application code.
- Per-key rate limits, spend quotas, and automated budget cutoff alerts.
- Community Edition open source core distributed under the AGPL-3.0 license.
Pricing
As listed on hyperionhq.co, checked 14 October 2026. Prices can change; see the site for current terms.
- Free: $0 for 10K monthly requests, 1 user seat, Redis caching, and automatic failover.
- Starter: $30/month for 30K monthly requests (overage at $8 per 100K), 5 user seats, and Qdrant semantic caching.
- Business: $99/month for 100K monthly requests, 50 user seats, 3-layer caching, and Jaeger tracing.
- Enterprise: Custom pricing for unlimited requests, VPC deployment, and dedicated support.
Facts from hyperionhq.co, checked 14 October 2026
- AI gateway written in Go providing an OpenAI-compatible routing interface.
- Supports 190+ endpoints across providers including OpenAI, Anthropic, and Google.
- Features multi-tier caching with Redis and Qdrant.
- Community Edition is open source under the AGPL-3.0 license, with cloud plans starting at $30/month.
Written from Hyperion's own website (Homepage, Pricing, Features, About). The maker can claim this listing to edit it.
Categories
Frequently asked questions
Which AI model providers does Hyperion support?
Hyperion supports providers including OpenAI, Anthropic, Google, Azure, Mistral, Groq, Together, Perplexity, Deepseek, Cohere, and Fireworks.
How does caching work in Hyperion?
Caching uses an L1 Redis exact-match layer, an L2 Qdrant semantic search layer, and an optional Layer 3 S3 archive.
Is an open source version of Hyperion available?
Yes. The Community Edition is open source and licensed under AGPL-3.0.
How is overage billed on Starter and Business plans?
Additional requests are billed at $8 per 100,000 requests up to 3 million requests per month.