FireWorldBench - Real-World-Aligned (Rendered Images): leaderboard

Metric: Completion accuracy (%; equal-weight mean of P1-P5 on the 714-item real-world-aligned subset, 26 event groups reconstructed from recorded indoor fire experiments; 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 September 2027. 10 models tracked.

Top models

#ModelScore
1GPT-5.6 Luna51.51
2Gemini 2.5 Flash47.65
3Qwen 3 VL 32B Instruct47.38
4Claude Haiku 4.545.95
5Qwen 3.5 Plus (2026-02-15)43.47
6Qwen 3 VL 235B A22B Instruct41.47
7InternVL3-78B38.47
8Qwen 3 VL 8B Instruct36.68
9InternVL3-38B36.62
10InternVL3-8B26.93

Interactive version: theaggregate.ai/benchmark?slug=fireworldbench-real-world-aligned-rendered-images · How It Works · Data refreshed daily, snapshot 2026-09-26.