See what you'd save on
document processing.
Put in your real volumes and see what moving off manual entry, AWS Textract, or Amazon Bedrock onto a self-hosted GPU pipeline would cost. Current public AWS rates, updating as you type.
Your volume
0 pages / month total
Your current method
Uses current AWS Textract per-page rates for the selected API.
Uses current Amazon Bedrock per-token rates for the selected model.
Horus self-hosted target
A GPU pipeline on your AWS account, the cheaper target you're comparing against.
Our estimate — adjust if you know your rate. This is the biggest driver of self-hosted cost, so the savings range below widens around it.
Blended rate: $0.81/hr · was $0.98
Spot prices vary by region and availability zone; Horus tunes the mix during a real engagement.
Your estimated savings
$0/yr
lower than your current method
Base estimate: $0/yr
$0/yr · $0.000/doc
$0–$0/yr · $0.000/doc · 0 instance(s)
What the Horus number includes
- GPU compute on your own AWS account — no per-page vendor fee
- Scales with instances as your volume grows
- Optional spot-instance mix for a further 60–90% cut
Live pricing was unavailable at build time; figures may be approximate.
Estimates based on public AWS pricing (Textract & Bedrock pulled 2026-11-02) and modeled self-hosted infrastructure assumptions. Actual costs vary with document complexity, accuracy targets, and workload. Source: live.