Neural inference,
built for production

Inference is a lightweight platform for running and serving open neural models at scale — with predictable latency, transparent pricing and no lock-in.

Get API keyLearn more

Low latency

Models served on dedicated inference nodes with sub-second cold start and batching tuned for throughput.

Model catalog

Curated open models with pinned versions, embeddings and chat endpoints behind one consistent API.

Simple pricing

Pay per request, no minimums. Usage-based billing with spend alerts and per-project budgets.

One API, every model

curl https://api.inference.example/v1/completions \
  -H "Authorization: Bearer $INFERENCE_KEY" \
  -d '{"model":"inference-1","prompt":"hello"}'