Skip to main content
Infyrence is a unified, OpenAI-compatible inference gateway. Call 200+ models from every major provider through a single endpoint and one API key, with automatic failover and edge-fast routing.
The gateway is a hosted commercial API. It is a separate product from the open-source infy framework, documented under Framework.

Why the gateway

One endpoint, one key

Reach OpenAI, Anthropic, Google, and more through a single base URL and one Infyrence API key. No per-provider SDKs or keys to juggle.

200+ models

Switch models by changing the model field. Call GET /v1/models for the live list available to your account.

Automatic failover

When an upstream provider is rate limited or briefly unavailable, the gateway retries and routes around it so your request still completes.

Edge-fast routing

Requests are routed at the edge for low-latency delivery to the nearest healthy provider.

OpenAI compatibility

The API mirrors the OpenAI Chat Completions API. To migrate an existing integration, point the base URL at the gateway and set model to any id from GET /v1/models. No other code changes are required.
Base URL
Because the API is OpenAI-compatible, most existing OpenAI SDKs, tools, and libraries work by only swapping the base URL and key.

Authentication

Authenticate every request with your Infyrence API key as a Bearer token.
Create and manage keys in your dashboard. A missing or invalid key returns 401 with an authentication_error.

Models

The model field accepts any id returned by GET /v1/models. Example ids include claude-sonnet-5, gpt-5.5, and gemini-3.1-pro. Always call the endpoint for the current list, since availability changes over time.
Each entry reports its id, owned_by, and capabilities (chat, completions, embeddings). See Models for details.

Billing

Billing is usage-based, per model. You pay for input and output tokens at each model’s published per-million-token rate. Every completion response includes a usage object with prompt_tokens, completion_tokens, and total_tokens. See the pricing page for current rates. If your account has no remaining balance for platform models, requests return 402 with an insufficient_credits error. Add funds, or use your own provider key (BYOK).

Bring your own key (BYOK)

Instead of drawing on your platform balance, you can supply your own provider key and pay that provider directly. Supported BYOK providers include openai, anthropic, groq, xai, deepseek, mistral, together, cerebras, alibaba, nvidia, and openrouter. You keep the same single Infyrence endpoint, key management, and OpenAI-compatible surface. See Pricing and usage for how BYOK is billed.

What you get

1

Point your SDK at the gateway

Set the base URL to https://api.infyrence.com/v1 and use your Infyrence API key.
2

Pick a model

Set model to any id from GET /v1/models.
3

Send requests

Call POST /v1/chat/completions. Set stream: true for incremental Server-Sent Events.

Next steps

Quickstart

Send your first request in a few minutes.

Models

Browse available models and their capabilities.

Pricing and usage

Understand per-model rates, usage, and BYOK.

API reference

Explore every endpoint, field, and error type.