AgroTools - Summary Accuracy: leaderboard

Step-by-step mode on AgroTools (2026), image-grounded agricultural questions (pest and disease diagnosis, counting and measurement, change detection, visualization) with 14 executable agricultural tools: given the gold history, the similarity of the final answer to the reference at gold answer steps, with penalties for numeric inconsistency and missing content, scaled to 0-100; higher is better.

Metric: Summary accuracy (0-100): normalized similarity of the final answer to the reference at gold answer steps, with penalties for numeric inconsistency and missing key content and zero when the model calls a tool instead of answering, in step-by-step mode on AgroTools (539 image-grounded agricultural queries with human-specified, human-audited reference tool trajectories over 14 executable tools; the model predicts the next step given the gold history), Lagent ReAct schema, temperature 0.1; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1InternVL3.5-14B33.96#1504
2InternVL3.5-8B33.72#1604
3Qwen 3 VL 8B Instruct28.24#1029
4Claude Sonnet 4.625.53#173
5GLM-4.6V22.48#710
6LLaVA-v1.6-Vicuna-13B21.83#5527
7Gemini 2.5 Pro21.01#276
8InternVL3-8B20.52#1782
9Seed 2.0 Pro17.52#222
10Qwen 3.5 9B16.7#828
11Qwen 3.5 35B A3B13.88#492
12GPT-5.410.93#135
13Janus-Pro-7B1.34#6857

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=agrotools-summary-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-21.