Leaderboard

What this leaderboard measures

The main track evaluates technical agronomic understanding: how reliably a chatbot or AI advisory service handles core technical concepts drawn from US land-grant extension publications. Version 0.5 does not cover field-level decisions such as seed or hybrid selection, input product choices, or recommendations conditioned on farm-specific data. Those are planned for future releases.

Results marked with an asterisk come from the post-cutoff subset — questions drawn from documents published after 30 September 2024, used to check whether a model's performance depends on having seen the source material in pretraining.

Results

How to read these scores

Each response is scored from 0 to 100 on four dimensions. The definitions below are the canonical ones and appear identically on /methodology and /glossary.

A high score means the model answered this benchmark's questions well. It is not a certification, a safety claim, or a statement about any tool's performance on questions outside the benchmark's scope.

Download the data

• CSV: /data/leaderboard-latest.csv (also at /data/leaderboard-YYYY-MM.csv)
• JSON: /data/leaderboard-latest.json

Both are advertised in the page's Dataset structured data so dataset-aware search surfaces can find them.

Submit a model

Proprietary and open-source systems are both welcome. See submit an AI system for evaluation.