Build with intelligence

The Vireonix API is free, requires no accounts and no API keys, and gives you access to language models through an OpenAI-compatible interface.

Make your first request Send a request. There is nothing to sign up for. Rate limits Fair use per IP so everyone stays free. Commercial resale Meter and sell Vireonix-powered AI inference.

Quickstart

Send a message with any HTTP client. No authentication header is required. The API accepts JSON and returns a standard chat completion object.

1. Send a request

curl https://vireonix.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      { "role": "user", "content": "Hello, Vireonix!" }
    ]
  }'

2. Use the response

The generated message is in choices[0].message.content.

{
  "id": "chatcmpl_vx_01H...",
  "object": "chat.completion",
  "model": "auto",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "Hello! What can I help you build?"
    },
    "finish_reason": "stop"
  }],
  "usage": { "prompt_tokens": 12, "completion_tokens": 9 }
}

Rate limits

Vireonix is free for everyone. Each IP receives hourly input and output budgets shown as a percentage remaining. Auto is reserved conservatively because the custom router may select a more capable upstream model:

Model	Burn rate	Input at 100%	Output at 100%
Auto	1x	20,000,000	200,000

Exceeding a limit returns HTTP 429 with error.type: rate_limit_exceeded. Response headers expose percentage remaining through X-RateLimit-Remaining-Input, X-RateLimit-Remaining-Output, and X-RateLimit-Remaining.

Check your current remaining budget anytime:

GET /api/limits

Models

auto is the only public model ID. For every request, Auto selects the best available model by considering task complexity, coding needs, Q&A quality, and response speed.

Model	Best for	Context
`auto`	Complex reasoning, coding, Q&A, and fast responses	Up to 1M tokens

Current model pool: MiniMax, Nemotron, GLM, Gemma, Llama, Qwen, GPT-OSS, and selected GPT models. Pool availability may change as models are evaluated and updated.

Messages

Use system messages for behavior, user for input, and prior assistant messages to keep context.

Streaming

Add stream: true to receive server-sent events as tokens are generated.

Errors

Use the HTTP status and error.type field. Retry 429 and 5xx with exponential backoff.

Chat completions

POST /v1/chat/completions (or local /api/chat)

Creates a model response from a list of conversation messages. No API key required.

List models

GET /v1/models

Returns every model currently available.

Limits status

GET /api/limits

Returns the input/output percentage remaining for your IP, plus model burn rates, for the current hour.

Reselling AI inference tokens

You may place Vireonix behind your own authenticated API or application, meter customer usage, and charge for Vireonix-powered AI inference tokens or credits. No upstream API key or prepaid wholesale balance is required.

Your reseller gateway must comply with hourly limits, enforce acceptable use, and disclose Vireonix data handling to customers. You provide retail authentication, metering, billing, refunds, and support. Public reseller access does not include an SLA or reserved capacity.

Read the AI token reseller guide and the controlling reseller terms before launching.