Pick a mode, not a model.

Quorum routes your prompt through a panel of frontier models, has them critique each other, and returns one synthesized answer — behind an API shaped like the one you already use.

Get a key → Read the docs

curl https://www.quorumllm.com/v1/chat/completions \
  -H "Authorization: Bearer $QUORUM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "QRUM:STAN", "messages": [{"role": "user", "content": "..."}]}'

OpenAI-compatible. Swap model and base_url — most SDK integrations just work.

Two modes today

QRUM:STAN

Engines at cost + platform fee + the Mode's fee

Three seats across Anthropic, Google, and OpenAI. Critique activates on harder questions, so a harder question deliberates more and costs more.

QRUM:EINS

Engines at cost + platform fee + the Mode's fee

Frontier-only seats. Critique runs on every question — Quorum's highest-quality deliberation.

How billing actually works

Prepaid balance, topped up via Stripe Checkout — no surprise invoices.
Every call's real cost is reserved before it runs and settled to the real amount after — you never pay for a call that failed.
Per-call cost is visible on your dashboard, not just a monthly total.
Each Mode's fee is published per 1M tokens processed, by tier, at GET /v1/models. The engines are billed at what their provider charges.

Know the price before you pay.

Send /v1/estimate the same prompt and model you'd send to /v1/chat/completions. You get back the real difficulty classification and a quote for the whole call: the engines, the platform fee and the Mode's fee, from the same classifier and billing formula production runs on. The bill follows what the call actually used, so it can differ from the quote.

curl https://www.quorumllm.com/v1/estimate \
  -H "Authorization: Bearer $QUORUM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "QRUM:STAN", "messages": [{"role": "user", "content": "..."}]}'

# → {"depth": "medium", "estimate": {"gateway_usd": …, "platform_fee_usd": …, "mode_fee_usd": …, "total_usd": …}, …}

Always free

$0 every time — call it as many times as you want while you model a workload.
Doesn't touch your balance or count against usage.
Same classifier, same pricing table as the real call — not a guess.

Bring your own key

Live and self-serve for Enterprise API organisations. Seat calls run on your own OpenAI, Google or Groq key, or through a Bedrock cross-account IAM role, and bill to your provider account. Orchestration — the judge and synthesis passes — stays on Quorum's keys.

The organisation owner adds keys under Usage.

Add a key in Usage →