Score is AI Inference Stack.Here is exactly what that means.
What you install, where it sits in your architecture, what hardware it uses, what connects to it, and how to buy. Five questions, five plain answers.
What software do I install?
Score SDK. Install it on your own hardware, in your cloud account, or in a data center; your applications reach it through an OpenAI-compatible API.
Where does it sit in my architecture?
Between your applications and your hardware. Your applications call an OpenAI-compatible endpoint that Score SDK exposes inside your environment.
What hardware does it use?
Runs on
- Intel Xeon
- AMD EPYC
- NVIDIA
In your account on
- AWS
- Google Cloud
- Azure
- Alibaba Cloud
What applications connect to it?
Anything that speaks the OpenAI-compatible API: coding agents, LangChain and LlamaIndex applications, internal tools and batch jobs. Change the base URL and keep the code.
Speaks to
- LangChain
- LlamaIndex
- Hugging Face
- Ollama
How do I buy?
Licensed against deployed capacity, sockets and nodes, not per token. It starts with a benchmark on your workload, then a proof of concept on one server with one model and one use case, then a license sized to what you deploy. Usage growing does not automatically grow the bill.
Benchmark your workload.
Send us a real workload. We run it on your models, your context lengths and your concurrency, and hand you the numbers either way.