Models

Heuristic Effort Units

Untangled does not bill per request, per token, or per page — those metrics are fragile. We bill per Heuristic Effort Unit (HEU), a single number that scales with how much actual work each call took. A short receipt costs less than a 40-page contract, and you can see exactly how much before each call is committed.

The HEU formula is internal and stable across model upgrades — it is not the underlying provider's token count. We absorb the variance so your bill doesn't spike when we change models.

The response includes the HEU charge for transparency:

{
  "vendor": "Acme Co.",
  "total": 1250.00,
  "_meta": {
    "heu": 12,
    "plan_remaining": 99988
  }
}

Plan ladder

• Free stops returning new responses once the bucket hits zero. There is no overage path.
• Paid plans keep returning responses past the included bucket — overage is billed monthly at the model rates above.
• Concurrent is the cap on simultaneous in-flight requests; exceeding returns 429 too_many_requests.

Buckets reset on the first of each month at 00:00 UTC.

Pre-flight HEU estimate

You can estimate cost before committing — pass ?estimate=true to any /v1/extract/* endpoint. The request returns the projected HEU charge without producing a structured response. Useful for batch jobs where you want to budget upfront.

curl "https://api.untangledapi.com/v1/extract/invoice?estimate=true" \
  -H "x-api-key: $UNTANGLED_KEY" \
  -H "content-type: application/json" \
  -d '{ "document": { "text": "..." } }'

{ "estimate": { "heu": 12 } }

Estimate calls do not consume your HEU bucket and are not rate-limited the same way live calls are.