ForecastBench

Based on the available evidence, ForecastBench is best adopted as a research benchmark rather than a product that directly delivers business forecasts. The GitHub repository describes it as a dynamic benchmark for LLM forecasting accuracy with contamination-free intent and human comparison groups. That supports its positioning, but it does not yet prove broad real-world adoption or benchmark dominance.

Its practical value is in giving researchers, evaluators, and agent builders a framework to measure forecasting performance over time, compare models against each other, and compare models against humans. It is not a general BI tool, and it is not a financial prediction service. A more accurate analogy is a public exam-and-scoring framework for studying forecasting ability.

On cost and adoption effort, the evidence is mostly the official GitHub repo. That is enough to support “open source,” but not enough to support claims about hosted pricing, API fees, or stable operating cost. Conservatively, the main cost is likely research and evaluation labor—dataset handling, experiment design, model runs, and reproducibility—rather than license fees.