Pricing Plans

Transparent Pricing Models

Pay for structured data volume. Choose a pre-built plan or adjust our dynamic token slider to fit your exact training budget.

Standard Ingestion

Standard

Standard web dataset compilation for pretraining 7B-13B parameter models.

$30,000/ contract

500M clean tokens
16 concurrent crawling threads
Rotating residential proxy exits
Hugging Face & pgvector sync (manual)

Dynamic Volume Plan

Scale Scraper

Scale Ingest

Adjust slider to compile larger dataset partitions.

$50,000/ contract

Ingestion Tokens:10B tokens / contract

Worker Concurrency:16 Concurrency Threads

Residential Proxies:1,500 Residential Nodes

Base Ingestion Fee:$35,000

Proxy Routing Allocation:$15,000

Auto PII scrub & deduplication
Hugging Face & Pinecone export sync
High-speed residential rotating proxies

Bespoke Pretraining

Foundation

Bespoke multi-trillion token datasets and dedicated regional crawling clusters.

$2,000,000+/ contract

1T+ high-quality pretraining tokens
Dedicated exit proxy subnets (unlimited)
Bespoke parsers & 24/7 Slack engineers
Real-time AWS S3 / Redshift streams

Technical Specification Matrix

Crawler Parameter	Standard	Scale Ingest	Foundation
Ingestion Dataset Volume	500M tokens	10B - 100B tokens	1T+ custom corpus
Concurrent Crawler Threads	16 threads	Up to 256 threads	Dedicated Cluster
IP Residential Proxy Pool	Shared Datacenter	Rotating Residential (15k+)	Dedicated Whitelist Nodes
MinHash LSH Deduplication	Standard clean	Fully automated v2	Custom Jaccard tuning
PII Anonymization scrubber	Basic regex	Automated regex/NLP	Custom Entity classifiers
Destination Sync Targets	Manual Download	Hugging Face, Milvus, pgvector	AWS S3, Redshift stream
CAPTCHA solver integration	Not available	CNN solve model inline	Bespoke solve routing
Support Response Latency	Email (48h)	Slack & Email (<12h)	Dedicated solutions engineer

Pricing FAQ

How are tokens calculated?

One token corresponds to approximately 4 characters of parsed markdown text or raw structured JSON outputs. Repetitive boilerplate code stripped by the LSH sweep does not count towards token usage.

Is there a daily scraping limit?

No. Your concurrent thread capacity determines how fast data is compiled. Free plan users are capped at 2 concurrency threads.

Can I bring my own proxies?

Yes. In the Scale plan and above, you can specify custom residential proxies. By default, Reworkd manages exit nodes automatically.

What formats do you support?

JSON, JSONL, Parquet, and CSV file formats. You can sync directly to Hugging Face, S3 buckets, or Pinecone databases.