Price PII detection before it hits your LLM
Amazon Comprehend Pricing Calculator
Estimate NLP units for entity detection, PII redaction, and custom classification — the preprocessing layer your Bedrock bill assumes is free.
- No signup
- Instant results
- ~3 min
Rates reviewed
Feature
us-east-1list rates, pricing as of 2026-10-17. Bedrock inference not included. Source: AWS public pricing.
Volume
3-unit (300 char) minimum per request.
Processing mode
$3/hr training.
$0.50/model/month.
Estimated Comprehend Cost
NLP processing
Total Monthly Cost
live resultAsync vs sync endpoint comparison unlock after email
Unlock Comprehend cost breakdown
See NLP unit count, free-tier credit, and Bedrock preprocessing context.
Full breakdown unlocked
Unit breakdown is now visible.
Want this applied to your real account?
FactualMinds pipelines PII redaction and Bedrock ingestion with cost guardrails for HIPAA and GenAI workloads.
Talk to GenAI Experts →Who This Tool Is For
ML engineers and compliance teams scrubbing documents before Bedrock ingestion or building non-LLM NLP pipelines.
Why We Built This Tool
Comprehend charges per 100-character unit with different rates for PII, custom models, and async batch vs sync real-time endpoints.
What Problem It Solves
- Unit minimum. Every request bills at least 3 units (300 characters) even for short text.
- Endpoint idle cost. Sync endpoints bill per second while running, regardless of request volume.
- No Bedrock overlap. This estimate excludes LLM token costs — model those separately.
See Generative AI on AWS or Bedrock pricing.
Frequently Asked Questions
How is a Comprehend NLP unit calculated?
One unit equals 100 characters, rounded up per request, with a 3-unit (300 character) minimum. A 550-character document uses 6 units.
Is PII detection priced differently from entity detection?
Detect PII uses the same $0.0001/unit rate as standard entity recognition. Contains PII is much cheaper at $0.000002/unit because it returns a boolean.
When is async batch cheaper than real-time?
Async batch suits bulk preprocessing pipelines. Sync endpoints win when you need sub-second latency and can keep utilization high enough to amortize idle time.
Do custom Comprehend endpoints run 24/7?
You pay from endpoint creation until deletion at $0.0005/second per inference unit — schedule endpoints down outside peak hours to cut cost.
Comprehend vs Macie for S3 PII — cost tradeoff?
Macie discovers PII in S3 buckets at scale; Comprehend redacts inline text before indexing. Use Macie for inventory and Comprehend for document pipelines.
What about toxicity detection and prompt safety classification?
Both bill per 100-character unit with a 3-unit minimum per request, at $0.0001/unit for the first 10M units — the same rate as standard NLP, so the Standard NLP option gives the cost. Toxicity detection is not in the free tier. Prompt safety classification, topic modeling, and event detection are no longer available to new customers.
Where do these prices come from?
Rates are AWS published us-east-1 list prices (Amazon Comprehend pricing), pricing as of 2026-10-17.
