FireWorldBench (Structured Records): leaderboard
Metric: Completion accuracy (%; equal-weight mean of the five physical-capability dimensions P1-P5 on the 8,360-item controlled-simulation subset; 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 2028. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Luna | 55.34 |
| 2 | DeepSeek V3.2 | 49.28 |
| 3 | Gemma 3 27B (IT) | 48.59 |
| 4 | Kimi K2 | 47.32 |
| 5 | Qwen 3.5 Plus (2026-02-15) | 47.25 |
| 6 | Gemini 2.5 Flash | 46.2 |
| 7 | Qwen 3 32B | 45.3 |
| 8 | Claude Haiku 4.5 | 44.78 |
| 9 | Qwen 3 235B A22B | 43.78 |
| 10 | Qwen 3 Next 80B A3B Instruct | 39.48 |
| 11 | GLM 4.5 Air | 39.23 |
| 12 | Llama 3.3 70B Instruct | 38.8 |
| 13 | Qwen 3 8B | 38.07 |
| 14 | Gemma 3 12B (IT) | 33.7 |
| 15 | Llama 3.1 8B Instruct | 31.64 |
Interactive version: theaggregate.ai/benchmark?slug=fireworldbench-structured-records · How It Works · Data refreshed daily, snapshot 2026-09-26.