Privacy Policy
This policy covers the data Benchwick collects across our website, sales channels, and inference API, including the prompts, inputs, and outputs you send to models we host.
Who we are
Benchwick LLC operates this website and the Benchwick inference service. For questions about this policy or to exercise a privacy right, contact us at sales@benchwick.com.
Data we collect
- Account and contact data — names, business email addresses, company details, and billing information you provide to us or through a reseller.
- Inference content — the prompts, messages, files, and parameters sent to the Benchwick inference API, and the model outputs returned to you.
- Usage and technical data — request identifiers, model identifiers, token counts, latency, error codes, and the originating IP address.
- Website data — standard server logs and analytics events collected when you browse this site.
How inference content is handled
Benchwick processes inference content only to serve the request and to bill for the tokens consumed. Content is held in volatile memory on the serving hosts for the duration of the request and for a short overlap window used to complete streaming responses.
- Retention. Prompt and completion bodies are retained for no longer than 30 days after the request, and only where request logging is enabled for support and debugging. After that period the records are deleted.
- Aggregate metrics. Token counts, latency, and error rates are retained indefinitely in aggregate form. These metrics contain no prompt or completion text.
- Training. Benchwick does not train models on your prompts or completions, and does not sell inference content to third parties.
Zero data retention
Benchwick does not currently operate a per-account zero data retention tier. If you require that request bodies are discarded immediately after the response is delivered, tell us before you start sending production traffic and we will confirm whether it is available for your account.
Why we process data
- To deliver inference responses and maintain service availability.
- To meter usage and issue invoices, including usage reports to our routing partners.
- To diagnose faults, prevent abuse, and secure the service.
- To respond to support requests and to comply with legal obligations.
Who we share data with
We share data with processors that operate the service on our behalf: infrastructure and hosting providers, our routing and billing layer, and email and support tooling. We may disclose data where required by law, or to protect the rights and safety of Benchwick, our customers, or the public. We do not sell personal data.
Where data is processed
Inference is served from hosts we operate in China, and service metadata is processed in the United States. If your workload requires a specific region, confirm it with us before sending production traffic.
Security
Access to inference hosts and logging is restricted to named operators, credentials are rotated, and traffic to our public endpoints is served over TLS. No system is perfectly secure; report a suspected issue to sales@benchwick.com.
Your choices
You may request access to, correction of, or deletion of personal data we hold about you. Because inference content is retained for a limited window, deletion requests for prompt or completion bodies should be sent within that window. Requests are handled at the contact address above.
Changes
We may update this policy as the service changes. The revision date below reflects the current version, and material changes will be announced to active accounts.
Last updated: November 17, 2026.
