pass.wafer.ai and an EU edge at pass-eu.wafer.ai. Customers who need EU data residency point their SDK at the EU hostname; everyone else uses the default US endpoint.
Endpoints
| Region | OpenAI-compatible | Anthropic-compatible |
|---|---|---|
| US (default) | https://pass.wafer.ai/v1 | https://pass.wafer.ai |
| EU | https://pass-eu.wafer.ai/v1 | https://pass-eu.wafer.ai |
Authentication
The samewfr_… Serverless API key works against both regions — there’s no separate “EU key” to provision. Auth, billing, quotas, and your account-level credit balance are shared across regions.
This means switching regions is a one-line config change:
Available Models in EU
The EU edge currently serves a curated subset of the catalog:| Model | Notes |
|---|---|
GLM-5.1 | 202K context, reasoning, ZDR |
Qwen3.5-397B-A17B | 262K context, vision, reasoning, ZDR |
Kimi-K2.6, MiniMax-M3, deepseek-v4-pro) returns model_not_found when sent to pass-eu.wafer.ai. Call GET https://pass-eu.wafer.ai/v1/models for live availability.
Data Residency
All inference onpass-eu.wafer.ai runs on EU-resident workers. Prompts, completions, and any expanded file frames stay within EU infrastructure for the duration of the request.
Zero Data Retention applies the same way it does on the US edge: send Wafer-ZDR: required to enforce no request-body retention. On the EU edge, ZDR is the default for the listed models — the header is still accepted for client-side certainty.
What’s Not Shared Across Regions
A few things are scoped to the region you uploaded to or operated against:- Files: a file uploaded to
pass.wafer.ai/v1/filesis not visible frompass-eu.wafer.ai. Re-upload to the EU edge if you need it there. - Metrics: the Serverless Metrics API and the dashboard
/usageview aggregate per-region. To see EU traffic, query the EU host the same way you query the US host. - Request IDs:
x-request-idvalues are unique per region.
Errors
If you hit a model that isn’t deployed in EU, you get the standardmodel_not_found with the live model list for the region: