Home

What AI AgriBench does

AI advisory tools for farming are multiplying. More than twenty commercial and open-source chatbots are already deployed in agriculture, alongside general-purpose assistants that farmers use anyway. Until AI AgriBench, no independent body evaluated whether any of them actually work.

The consortium builds benchmark datasets from authoritative extension publications, peer-reviewed research and real farmer–expert interactions, scores AI models and commercial services against them, and publishes the results openly so that farmers, agronomists, agtech companies, funders and policymakers have an objective basis for deciding which tools to trust.

AI AgriBench evaluates natural-language agronomy answers; it is distinct from similarly named multimodal or computer-vision benchmarks.

Expert-grounded benchmarks

Benchmark datasets are drawn from more than 40,000 extension publications and research papers from US land-grant universities, then reviewed by 23 agronomists. Every response is scored from 0 to 100 on accuracy, completeness, relevance and conciseness — never a single black-box number.

Academic and industry depth

Partners span AI research, crop science, extension and production deployments. Twelve founding members sit alongside a growing membership across industry, academia and the nonprofit sector, with governance shared rather than held by any one vendor.

Transparent results

Model scores are filterable by agronomic topic category and downloadable as CSV and JSON. Every release is archived at a permanent URL so a result can be cited years later exactly as it was published.