Build on Tempest.

What you are calling

Failure Tempest 1 fuses four frontier models — Claude Fable 5.5, Claude Opus 5.5, GPT-6 Astra 6.2 and Gemini 4 Argon — into one answer. Easy prompts stream straight from the best-fit model. Hard prompts get a pre-answer brief, several drafts written under different lenses, a fusion step that settles disagreements by checking, and an audit by a model that did not write the answer. You see one response. You pay for the tokens you send and receive, never for the drafts behind it.

Read API fusion for the full pipeline.

The basics

Your first request

curl https://api.failure.fail/v1/chat/completions \
  -H "Authorization: Bearer $FAILURE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "failure-tempest-1",
    "messages": [
      {"role": "system", "content": "You are a careful assistant."},
      {"role": "user", "content": "Is 2^61 - 1 prime? Show the check."}
    ]
  }'

Full walkthrough, including keys and the Python and TypeScript SDKs, in the quickstart.

Where to go next

Getting help

API questions go to api@failure.fail. A person reads every message. For anything else, use the contact page.