FireWorldBench - Cross-Field Coupling Understanding (Rendered Images): leaderboard

Metric: Completion accuracy (%; P3, cross-field coupling understanding items of the controlled-simulation subset; image-observation track (physics-rendered field visualizations and 3D scene renderings); each cell is the mean of the matched choice-format score (Jaccard overlap of selected and gold option sets) and open-report score (exact agreement of required structured slots); deterministic evaluator, no LLM judge; failed or unparseable responses score 0). Source: arxiv.org. Saturation forecast: Around 2029. 10 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash45.86
2GPT-5.6 Luna44.09
3Qwen 3 VL 32B Instruct41.5
4Qwen 3.5 Plus (2026-02-15)40.72
5Claude Haiku 4.540.49
6InternVL3-38B37.95
7Qwen 3 VL 235B A22B Instruct36.66
8InternVL3-78B35.84
9InternVL3-8B34.08
10Qwen 3 VL 8B Instruct32.45

Interactive version: theaggregate.ai/benchmark?slug=fireworldbench-cross-field-coupling-understanding-rendered-images · How It Works · Data refreshed daily, snapshot 2026-09-26.