FireWorldBench (Rendered Images): leaderboard
Metric: Completion accuracy (%; equal-weight mean of the five physical-capability dimensions P1-P5 on the 8,360-item 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 2028. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Flash | 44.83 |
| 2 | GPT-5.6 Luna | 42.99 |
| 3 | Qwen 3 VL 32B Instruct | 42.55 |
| 4 | Claude Haiku 4.5 | 39.51 |
| 5 | Qwen 3.5 Plus (2026-02-15) | 39.4 |
| 6 | InternVL3-78B | 38.72 |
| 7 | InternVL3-38B | 34.81 |
| 8 | Qwen 3 VL 235B A22B Instruct | 34.02 |
| 9 | Qwen 3 VL 8B Instruct | 29.41 |
| 10 | InternVL3-8B | 27.77 |
Interactive version: theaggregate.ai/benchmark?slug=fireworldbench-rendered-images · How It Works · Data refreshed daily, snapshot 2026-09-26.