OpenBenchmarks helps AI agents compare SaaS APIs
OpenFunnel founders Fenil and Aditya launched OpenBenchmarks. It gives AI agents reproducible benchmarks for comparing SaaS APIs and deciding whether to buy a service or build around another option. The first focus is sales and marketing APIs.
The core bet is that B2B software choices will increasingly happen inside using , not only through human research and sales material. OpenBenchmarks tries to replace vendor claims and search rankings with open data, public scoring rules, and reproducible tests. Each result includes the actual HTTP request and response, plus the prompt and response from the LLM judge.
One live benchmark compares lookalike-company tools: 24 seed companies, 4 vendors, and 100 returned companies per vendor, with the judge marking which results are useful. The top Precision@100 score is 57%, and the site also shows and benchmarks with speed and word error measurements.
Key points
- OpenBenchmarks is an site for agents comparing SaaS APIs.
- The first area is sales and marketing APIs, including lookalike-company tools.
- Results include the HTTP request, HTTP response, LLM judge prompt, and LLM judge response.
- The lookalike benchmark compares 24 seed companies across 4 vendors, with 100 results per vendor.
- The best Precision@100 score shown is 57%, and voice benchmarks compare speed and .