Supported providers
You can configure an upstream for 10 built-in providers and 25 OpenAI-compatible vendors, and for any other OpenAI-compatible endpoint through Custom OpenAI-compatible. Each row is a provider you can configure an upstream for. For how each provider family authenticates, see Supported providers in Providers; for a vendor's endpoints, pinned hosts, and quota signals, see the vendor catalog.
Providers
| Provider | Id | Kind | Tier | Wire formats | Authentication | Zero data retention | No prompt training | Jurisdiction | Featured |
|---|---|---|---|---|---|---|---|---|---|
| Anthropic | anthropic | Built-in | first-class | Anthropic Messages | API key | — | Yes | US | Yes |
| OpenAI (API key) | openai | Built-in | first-class | OpenAI Chat Completions, OpenAI Responses | API key | — | Yes | US | Yes |
| Azure OpenAI | azure (alias azure-openai) | Built-in | first-class | OpenAI Chat Completions, OpenAI Responses | API key or managed identity | Yes | Yes | US | Yes |
| AWS Bedrock (OpenAI-compatible models) | aws-bedrock (alias bedrock) | Built-in | first-class | OpenAI Chat Completions | AWS credentials | Yes | Yes | US | Yes |
| OpenRouter (free and paid models) | openrouter | Vendor | first-class | Anthropic Messages, OpenAI Chat Completions | API key | — | — | — | Yes |
| Google Vertex | google-vertex (alias vertex) | Built-in | first-class | Gemini generateContent | Google Application Default Credentials | Yes | Yes | US | Yes |
| Alibaba Cloud Model Studio (Qwen) | alibaba | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| AWS Bedrock (Anthropic Claude — Sonnet, Haiku, Opus) | aws-bedrock-anthropic | Built-in | first-class | Anthropic Messages | AWS credentials | Yes | Yes | US | — |
| Baseten Model APIs | baseten | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| Cerebras | cerebras | Vendor | first-class | OpenAI Chat Completions | API key | — | — | US | — |
| Cohere | cohere | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| DeepInfra | deepinfra | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| DeepSeek | deepseek | Vendor | first-class | Anthropic Messages, OpenAI Chat Completions | API key | — | — | CN | — |
| Fireworks AI | fireworks | Vendor | first-class | Anthropic Messages, OpenAI Chat Completions | API key | — | — | US | — |
| GitHub Copilot | github-copilot | Built-in | first-class | OpenAI Chat Completions | Each developer's own sign-in | — | — | US | — |
| Google Gemini | gemini | Vendor | first-class | OpenAI Chat Completions | API key | — | — | — | — |
| Google Vertex (Anthropic Claude) | google-vertex-anthropic | Built-in | first-class | Anthropic Messages | Google Application Default Credentials | Yes | Yes | US | — |
| Groq | groq | Vendor | first-class | OpenAI Chat Completions | API key | — | — | US | — |
| Hugging Face Inference Providers | huggingface | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| LiteLLM | litellm | Built-in | first-class | OpenAI Chat Completions | Optional API key | — | — | — | — |
| MiniMax | minimax | Vendor | community | Anthropic Messages, OpenAI Chat Completions | API key | — | — | — | — |
| Mistral AI | mistral | Vendor | first-class | OpenAI Chat Completions | API key | — | — | EU | — |
| Moonshot (Kimi) | moonshot | Vendor | first-class | Anthropic Messages, OpenAI Chat Completions | API key | — | — | CN | — |
| Nebius Token Factory | nebius | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| Novita AI | novita | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| NVIDIA NIM | nvidia-nim | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| Ollama | ollama | Vendor | local | Anthropic Messages, OpenAI Chat Completions | None | — | — | — | — |
| OpenAI (ChatGPT OAuth — Codex) | chatgpt-oauth | Built-in | first-class | OpenAI Responses | Each developer's own sign-in | — | — | US | — |
| Perplexity | perplexity | Vendor | first-class | OpenAI Chat Completions | API key | — | — | US | — |
| SambaNova Cloud | sambanova | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| Sarvam AI | sarvam | Vendor | community | OpenAI Chat Completions | API key | — | — | — | — |
| Together AI | together | Vendor | first-class | OpenAI Chat Completions | API key | — | — | US | — |
| vLLM | vllm | Vendor | local | Anthropic Messages, OpenAI Chat Completions | Optional API key | — | — | — | — |
| xAI | xai | Vendor | first-class | OpenAI Chat Completions | API key | — | — | US | — |
| Z.ai (GLM) | zai | Vendor | first-class | Anthropic Messages, OpenAI Chat Completions | API key | — | — | CN | — |
| Custom OpenAI-compatible | custom | Custom | custom | OpenAI Chat Completions | API key | — | — | — | — |
LiteLLM is configured in the config file only; neither the setup wizard nor Settings → Providers offers it.
Columns
- Kind: Built-in for a provider AIControls handles natively, Vendor for an entry in the vendor catalog, and Custom for any other OpenAI-compatible endpoint. A Custom upstream's provider name is
custom:<host>, such ascustom:llm.example.com. - Tier: first-class is every built-in provider, and each hosted vendor with tested endpoints, pricing, and, where the vendor reports them, its own credit and quota signals. community vendors have a complete catalog entry but aren't tested on a schedule. local is a model server you run, such as Ollama or vLLM, at any URL. custom takes any URL. See Tiers.
- Wire formats: the request formats the provider serves. A client that sends another format is translated where AIControls supports it; see Cross-format routing.
- Authentication: what an upstream for the provider needs. Each developer's own sign-in means each developer's AI tool sends its own credential, so no key is stored.
- Zero data retention, No prompt training, Jurisdiction: the provider's default data-handling guarantees, which data-handling policies check. They are defaults, not facts about your account: the jurisdiction of an Azure resource or an AWS region is where you deployed it. Set the upstream's capabilities to match its region and the terms of your own agreement with the provider. A Google Vertex upstream takes its jurisdiction from the region in its URL; see Jurisdiction of a Google Vertex upstream. A dash means the provider makes no such guarantee by default, or, for Jurisdiction, that it has no default jurisdiction.
- Featured: the provider is marked
featuredin the provider catalog API, which lists featured providers first.