Upstream Providers
Five infrastructure providers (Azure, Bedrock, Vertex, Bailian, Volcengine) power the 200+-model catalog with geographic diversity and cross-vendor redundancy.
Provider Overview
| Provider | Models Routed | Regions | Typical Latency |
|---|---|---|---|
| Azure (OpenAI + MaaS) | ~30 | Global | 80-120ms |
| Amazon Bedrock | ~50 | US multi-region | 100-150ms |
| Google Vertex AI | ~18 | US | 90-130ms |
| Alibaba Bailian (DashScope) | ~45 | Asia-Pacific | 50-80ms |
| Volcengine (ByteDance) | 3 | Asia-Pacific | 50-80ms |
Next Steps
- Provider Routing — How the gateway selects providers for each request
- Failover — Automatic rerouting when a provider is down
- Circuit Breaker — Health monitoring and recovery for upstream providers