Inferock Bench is an open-source developer tool designed to bring transparency and independent auditing to large language model (LLM) API consumption. Operating as a local proxy, it sits directly between your application code and your LLM service providers to monitor and log backend activity.

By proxying calls formatted for major providers—including OpenAI, Anthropic, Gemini, and OpenRouter—Inferock Bench captures key metrics on a per-call basis. It records token usage, tracks request failures, and monitors retries. Using this telemetry, it generates independent receipts that show what was billed compared to actual consumption, allowing developers to detect discrepancies and see exact overpayment figures.

Pros of Inferock Bench

Generates independent receipts to audit LLM API billing accuracy
Operates locally as a proxy, maintaining control over network traffic
Captures detailed per-call metrics including token usage, failures, and retries
Supports popular API formats for OpenAI, Anthropic, Gemini, and OpenRouter
Open-source software hosted on GitHub

Cons of Inferock Bench

Requires local proxy configuration within your application infrastructure
Specific to supported provider call formats (OpenAI, Anthropic, Gemini, OpenRouter)

Frequently Asked Questions

What is Inferock Bench?

Inferock Bench is an open-source local proxy tool that sits between your application and LLM providers. It monitors API traffic to generate independent receipts showing your actual token usage, failures, retries, and potential overpayments.

Which LLM providers does Inferock Bench support?

Inferock Bench supports API calls structured for OpenAI, Anthropic, Gemini, and OpenRouter.

How does Inferock Bench calculate billing discrepancies?

By acting as a proxy, it captures the exact token count, failure rates, and retries for every call made by your application, comparing actual usage against provider charges.