FireWorldBench - Real-World-Aligned (Structured Records): 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; structured-observation track (textual sensor traces and physical records); 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 December 2027. 15 models tracked.

Top models

#ModelScore
1GPT-5.6 Luna60.16
2Qwen 3.5 Plus (2026-02-15)55.35
3Claude Haiku 4.555.05
4Gemini 2.5 Flash54.35
5Llama 3.3 70B Instruct51.2
6GLM 4.5 Air50.55
7Kimi K249.54
8DeepSeek V3.249.24
9Qwen 3 Next 80B A3B Instruct48.98
10Qwen 3 32B47.64
11Gemma 3 27B (IT)44.83
12Qwen 3 235B A22B44.47
13Gemma 3 12B (IT)41.25
14Qwen 3 8B39.09
15Llama 3.1 8B Instruct32.54

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