Open by default
The open-weight ecosystem is where the most exciting models ship. We build infrastructure that lets anyone serve them, with no closed walled garden.
About
The infrastructure layer for the open ecosystem.
Open Scale is an independent inference cloud. We exist to make open-weight models fast and reliable to run, behind a single OpenAI-compatible API.
Our mission
Frontier capability is moving into open weights at an unprecedented pace. But open models only matter if anyone can actually run them at scale. We build the serving layer: the GPUs, the routing, the caching, the billing. So that developers and companies can use these models the same way they use any other API.
What we're building
- OpenAI-compatible API for open-weight models
- Hardware-efficient inference on optimized vLLM
- Transparent per-token pricing, no markup
- Prefix caching for lower cost and faster returns
- Zero-data-retention serving path
Principles
How we operate
No markup, ever
Our pricing reflects the silicon. What you see on OpenRouter is what we charge.
Your data is yours
We don't train on prompts and we don't sell them. A zero-data-retention path is available for workloads that require it.
Built for routing
We optimize for the metrics routers actually measure: uptime, first-token latency, and throughput, so your traffic lands where it should.
Let's route some traffic through us.
Whether you're evaluating, integrating, or want to talk enterprise, we're easy to reach.