Loading the catalogue…
Google's fast, cost-focused Gemini 3.5 model with a million-token context window, mandatory configurable reasoning, and multimodal text, image, audio, and video input.
Same model, one lane, your call. Sovereign first, then European-hosted, then routed (typically non-EU). Default endpoint per lane is highlighted.
| Provider | Input / M | Output / M | TTFT | Tok/s | Reliability | Status |
|---|---|---|---|---|---|---|
GoogleDefault | €0.26 | €2.15 | 1.49 s | 4 | 100.0% | Active |
Pick this model first.
Works — usually a fine pick.
Other models will serve you better.
Fit assertions set by the curator pipeline. 23 use cases mapped.
Model catalogue
Stav speaks the same chat-completions API you already use. Set STAV_API_KEY and the curl below works. SDKs (openai-python, langchain, litellm) all work with the same base URL.
curl https://api.stav.ai/v1/chat/completions \
-H "Authorization: Bearer $STAV_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-3.5-flash-lite",
"messages": [{"role": "user", "content": "Hello!"}]
}'Prefer the Anthropic SDK contract? The same model is also reachable at POST /v1/messages — Stav's gateway translates between Anthropic-shape and the underlying provider for non-Claude models.
Sovereign-EU inference by default; the Smart Router lifecycle penalty drains traffic away from deprecating endpoints automatically. Add X-Stav-App-Id: <uuid> to attribute requests to a registered app.
Inference runs in US on google. The model was created in US as well. Both training and inference fall under US compelled-disclosure regimes (CLOUD Act / FISA 702). Not appropriate for EU regulated-sector workloads without explicit legal sign-off.
Inference and (typically) training both happen under US jurisdiction — both CLOUD Act and FISA Section 702 reach. EU regulated-sector customers cannot use this combination for sensitive data without a derogation / case-by-case legal review.