Baba Is AI
Models see a puzzle and write a plan to reach the winning object. Some puzzles require them to push words together to change the rules.
The original authors scored exact plan matches. The BALROG team later measured progress on 40 levels with seven language models and six vision models.
- Made by
- MIT and University of Virginia
- Released
- July 2024
- Kind
- Puzzles and chess
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Environment
- Baba Is You
- Items from
- Written by experts
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
-
Gemini 1.5 Pro Sep 2024 #157 on the capability index
- Last model added
- March 2025
- Scores refreshed
- November 20th, 2026
Llama 3.2 90B leads with 43.9% among models released by March 2025
Best score per model from BALROG’s leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
The newest model with a score here came out in March 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1
Llama 3.2 90B LLM 43.9%
- 2
Llama 3.1-70B LLM 40%
- 3 GPT-4o Mar 2025, LLM 33.7%
- 4
Gemini 1.5 Pro Sep 2024, LLM 32%
- 5 GPT-4o mini VLM 16.4%
- 6
Llama 3.2 11B LLM 15.6%
- 7
Gemini 1.5 Flash Sep 2024, LLM 12.8%
Source: BALROG’s leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 43.9% | LLM | Sep 2024 | Official leaderboard |
| 2 | | 40% | LLM | Jul 2024 | Official leaderboard |
| 3 | GPT-4o Mar 2025 | 33.7% | LLM | Mar 2025 | Official leaderboard |
| 4 | | 32% | LLM | Sep 2024 | Official leaderboard |
| 5 | GPT-4o mini | 16.4% | VLM | Jul 2024 | Official leaderboard |
| 6 | | 15.6% | LLM | Sep 2024 | Official leaderboard |
| 7 | | 12.8% | LLM | Aug 2024 | Official leaderboard |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: BALROG’s leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| LLM | 43.9% | ||||
| LLM | 40% | ||||
| GPT-4o Mar 2025 | LLM | 33.7% | |||
| LLM | 32% | ||||
| VLM | 31.4% | ||||
| VLM | 21.9% | ||||
| GPT-4o Mar 2025 | VLM | 18.6% | |||
| GPT-4o mini | VLM | 16.4% | |||
| GPT-4o mini | LLM | 15.6% | |||
| LLM | 15.6% | ||||
| LLM | 12.8% | ||||
| VLM | 8.30% | ||||
| VLM | 5.76% |
3 models have held number one on BALROG’s leaderboard since its release in July 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: BALROG’s leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: BALROG’s leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Games and long tasks
- ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6.2 Astra 100%
- Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games.
Gemini 2.5 Pro 3.90
- MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks.
Seed2.0 Pro 49%
- TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress.
Llama 4 Scout 7.00
- GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
- Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 20 months ago, in March 2025, for GPT-4o. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (7 entries).
- Pass Models scored 7 models we track.
- Warning Newest score The newest came 20 months ago, in March 2025, for GPT-4o.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 1.5 Pro (#157).
- Pass Headroom The best is 43.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on BALROG: 13 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.