Skip to content
NewTry RouterPlus with free creditsTry playground →

Inference providers

16 providers serving 43 models on one key. 24 models are served by more than one provider, so a failed provider has a fallback.

  • 16Providers
  • 43Models
  • 10Labs
  • 24Models on more than one provider

All providers

ProviderModelsLabsMax context
DeepInfra17DeepSeek, Google, MiniMax, Moonshot AI, OpenAI, Qwen, Z.ai1.05M
Inference.net17DeepSeek, Moonshot AI, Qwen, Z.ai1.05M
StreamLake16DeepSeek, MiniMax, Moonshot AI, Qwen, Z.ai1.05M
Doubleword15DeepSeek, Google, Meta, Moonshot AI, Qwen, Z.ai1.05M
Nebius13DeepSeek, Meta, MiniMax, Moonshot AI, Qwen, Z.ai1.05M
Phala12DeepSeek, Moonshot AI, OpenAI, Qwen, Z.ai1.05M
Novita AI11DeepSeek, Google, Moonshot AI, Qwen, Z.ai1.05M
Chutes10DeepSeek, Google, Moonshot AI, Qwen, Z.ai1.05M
Sail Research8DeepSeek, Google, Moonshot AI, Z.ai1.05M
Wafer8DeepSeek, Moonshot AI, Qwen, Z.ai1.05M
Morph5DeepSeek, Moonshot AI, Z.ai1.05M
RunInfra5DeepSeek, Qwen, Z.ai1.05M
Munito AI3DeepSeek, Qwen, Z.ai1.05M
Particle AI3DeepSeek, Particle AI, Z.ai1.05M
Mistral AI1Z.ai1.05M
Paper Instruments1Paper Instruments262K

Open text models in the RouterPlus catalog, as of November 5, 2026.

How a request picks a provider

  1. Match the model

    Keep the providers that list the exact model id you asked for. No fuzzy matching and no silent substitute.

  2. Drop resting providers

    A provider cooling down after failures sits out until it recovers.

  3. Order the rest

    By each provider's live listed price, unless provider.order or your key's routing policy sets the order. The first to send output serves you.

Set the order yourself

OpenAI-compatible
curl https://api.routerplus.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.3-flash",
    "messages": [{ "role": "user", "content": "Hello" }],
    "provider": { "order": ["morphllm", "deepinfra-chat"] }
  }'

only limits the request to the providers you name, ignore skips them, and POST /v1/route shows the plan without sending the request. Routing docs

Reserved capacity

Tell us the model, throughput and region you need. RouterPlus shops the requirement across its provider network and replies with committed capacity, a better quote or a backup provider.

Send a capacity request

FAQ

What is an inference provider?

A company that runs model weights on its own hardware and sells access by the token. RouterPlus puts 16 of them behind one key, so you call a model and the gateway picks a provider that serves it.

How does RouterPlus pick a provider?

It keeps the providers that serve the exact model you named, drops any that are resting after failures, and orders the rest by live listed price. provider.order or a routing policy on your key changes that order. POST /v1/route shows the plan without sending the request.

Can I choose the provider?

Yes. Send a provider object with the request: order tries the providers you name first and keeps the rest as fallbacks, only limits the request to the providers you name, and ignore skips providers. The x-tm-provider response header names the provider that served you.

Do providers charge different prices?

Each provider sets its own price, and you pay the price of the provider that served the request. Every billed response carries usage.cost in USD, and each provider page lists that provider's price per model.

What happens when a provider goes down?

If it fails before any output reaches you, the gateway moves to the next provider for that model with no backoff. Once output starts, the request stays with that provider. A request limited with only has no fallback outside the providers you named.

Do providers keep my prompts?

Each provider declares its own prompt retention in the catalog, and each provider page shows it. RouterPlus billing keeps metadata only; requests on shared endpoints go to an encrypted request log for up to 30 days unless your organization opts out.

Every provider, one key