RebusBench (1-Shot) - GPT-4o Judge: leaderboard

RebusBench (2026) rebus puzzles scored by a GPT-4o semantic judge that rates the answer's similarity to the gold idiom from 0 to 1 (shown times 100), with one solved example in the prompt; higher is better.

Metric: GPT-4o semantic judge score (0-1, times 100) rating the similarity of the predicted answer to the gold answer, averaged over the 1,164 rebus puzzles of RebusBench, one in-context example in the text prompt; higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 VL 8B16.61#917
2Qwen 3 VL 4B16.5#1151
3LLaVA-v1.5-7B14.91#7630
4Qwen 2.5 VL 32B Instruct14.78#1062
5InternVL3.5-30B-A3B14.3#1267
6Qwen 2.5 VL 7B Instruct14.13#1902
7InternVL3.5-8B13.97#1604
8InternVL3.5-4B13.83#1930
9Qwen 2.5 VL 3B Instruct13.51#2713

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=rebusbench-1-shot-gpt-4o-judge · How It Works · Data refreshed daily, snapshot 2026-11-21.