# Model Providers

Route model traffic through Blekline **before** it reaches upstream APIs. Use [Ingress proxy](/docs/api/ingress-proxy) with SDK base-URL swap.

## Comparison

| Provider | Category | Doc |
|----------|----------|-----|
| [Azure OpenAI](/docs/integrations/azure-openai-stack) | Enterprise | EU residency, BYOK |
| [AWS Bedrock](/docs/integrations/aws-bedrock-stack) | Enterprise | IAM, VPC sidecar |
| [Google Vertex](/docs/integrations/google-vertex-stack) | Enterprise | GCP credentials |
| [OpenRouter](/docs/integrations/openrouter-stack) | Router | Multi-model dev |
| [Together](/docs/integrations/together-stack) | Router | Open models |
| [Fireworks](/docs/integrations/fireworks-stack) | Router | Fast inference |
| [Mistral](/docs/integrations/mistral-stack) | Specialty | Mistral API |
| [Cohere](/docs/integrations/cohere-stack) | Specialty | SDK mask path |
| [Replicate](/docs/integrations/replicate-stack) | Specialty | SDK mask path |
| [Qwen](/docs/integrations/qwen-stack) | Router | OpenRouter / DashScope |

## Universal ingress pattern

```python
from openai import OpenAI
client = OpenAI(
    base_url="https://app.blekline.com/api/ingress/v1",
    api_key="blw_...",
    default_headers={
        "x-blekline-workspace-token": "blw_...",
        "x-blekline-client-surface": "sdk",
    },
)
```

Set server `OPENAI_API_BASE` to your provider endpoint. Blekline masks user messages inline before forwarding.

---

**Next steps:** [Ingress proxy](/docs/api/ingress-proxy) · [Deployment](/docs/enterprise/deployment) · [Trust boundaries](/docs/security/trust-boundaries) · [Open workspace](https://app.blekline.com)