Provider Pinning & Max Savings

By default Qubax routes each request to the best available offer for the requested model. Power users can go further: pin routing to specific upstream providers, enforce a minimum savings floor, or register your own provider key as a priority route.

Provider pinning

Pass a provider field in the chat completion body — a single provider or an allow-list array. The request only routes to offers from the providers you list; everything else is excluded even if cheaper.

curl https://api.qubax.ai/v1/chat/completions \
  -H "Authorization: Bearer qbx_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3",
    "messages": [{"role": "user", "content": "Hello!"}],
    "provider": "zai"
  }'

Allow-list form — any of these providers is eligible:

{
  "model": "glm-5.3",
  "messages": [{"role": "user", "content": "Hello!"}],
  "provider": ["zai", "openrouter", "anthropic"]
}

Max Savings — minimum-discount routing

Set min_discount (1–100) in the request body to only accept offers at least that far below OpenRouter pricing. If nothing qualifies, the request fails with minimum_discount_not_met instead of silently paying more. You can also set a default per API key on the API keys page.

{
  "model": "claude-opus-4.6",
  "messages": [{"role": "user", "content": "Hello!"}],
  "min_discount": 30
}

Anthropic-compatible endpoint

Claude Code and the Anthropic SDK work on any Qubax model — not just Claude — through the Anthropic Messages skin:

export ANTHROPIC_BASE_URL="https://api.qubax.ai"
export ANTHROPIC_API_KEY="qbx_..."

# Claude Code now runs on any Qubax model

The endpoint accepts standard Anthropic POST /v1/messages payloads (system blocks, content blocks, max_tokens) and returns Anthropic-format responses and SSE events.