About API Benchmarks

API Benchmarks was founded in 2025 by a group of infrastructure engineers who were tired of marketing claims with no data to back them up. We run controlled, repeatable benchmarks across every major AI API provider and publish the raw results.

Our methodology: same prompts, same network conditions, same hardware. Every test runs 100 times, we discard the top and bottom 10, and report the average of the middle 80. We subtract network latency by measuring empty ping to the API server before each test run.

We are not affiliated with any AI provider. All tests use our own credits purchased at standard pricing.