Every frontier model, on one credit account

Pick by job

Every model

Prices are the provider’s rate per 1M tokens; Vaaya adds 3%. Pass the id as model.

Don’t know which model? Let your code decide

Send the prompt to Jev first. It labels how hard the task is, with a confidence, and your code maps the label to a lane: a cheap model for simple asks, a frontier model for hard ones, and a step up whenever confidence is low. In a published 50-prompt test, that routing cut usage cost 34.7% at a 34% slower p95.

One call, one receipt

The router speaks the OpenAI chat format, so any OpenAI SDK works with base_url set to https://vaaya.ai/api/llm/v1.

curl https://vaaya.ai/api/llm/v1/chat/completions \
  -H "Authorization: Bearer $VAAYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "anthropic/claude-sonnet-5",
       "messages": [{"role": "user", "content": "Summarize this changelog in three bullets"}]}'

The response carries the charge. usage.cost is what your credit pays for this call, in USD, with Vaaya’s 3% included; the same call lands in your transaction history with its model and tokens.

{
  "model": "anthropic/claude-sonnet-5",
  "choices": [{ "message": { "role": "assistant", "content": "…" } }],
  "usage": {
    "prompt_tokens": 1200,
    "completion_tokens": 180,
    "cost": 0.004326
  }
}

Token counts are illustrative; the cost shown is Claude Sonnet 5’s rate applied to them.

Why not a raw provider key

Questions