Your first call
Point any OpenAI SDK at the Celeris base URL, or curl it directly. Full walkthrough in the Quickstart.
curl https://inference.cloud.celeris.ai/celeris-1/v1/chat/completions \
-H "Authorization: Bearer $CELERIS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "celeris-1",
"messages": [{"role": "user", "content": "Reply with one word: fast or slow?"}],
"max_tokens": 4, "temperature": 0}'
OpenAI-compatible
The chat-completions API you already use, at a different base URL. Official SDKs work unmodified — swap the endpoint and go.
Making requests →Built for agents
Classify, extract, judge, rewrite: diffusion decoding lands whole short answers in one burst, so hot-path calls stop dominating your agent's latency budget.
Agentic workloads →Latency you can audit
Every response carries a Server-Timing breakdown of processing, queueing, and inference — measure us, don't trust us.