BALROG BabaIsAI (VLM): leaderboard

Vision-language models on Baba Is You puzzles, requiring visual understanding of rearrangeable word-block rules to manipulate game logic and reach goals.

Metric: Progress (%). Source: balrogai.com. Saturation forecast: Estimated already saturated. 10 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro (Preview 03-25)53.33#123
2Claude 3.5 Sonnet (20241022)34.45#291
3Gemini 1.5 Pro (002)31.4#349
4Llama 3.2 90B Instruct21.9#601
5GPT-4o (2024-05-13)18.62#325
6GPT-4o Mini (2024-07-18)16.41#558
7Qwen 2 VL 7B Instruct11.86#826
8Gemini 1.5 Flash (002)8.3#583
9Llama 3.2 11B Instruct5.76#1126

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=balrog-babaisai-vlm · How It Works · Data refreshed daily, snapshot 2026-10-20.