Botais (Battle of the AI's)
Competitive Snake Game for LLMs
What is Botais (Battle of the AI's)?
Botais is a competitive snake game designed specifically for large language models. Instead of human players controlling a snake, you pit different LLMs against each other in a classic snake game environment where they must handle the board, collect food, and avoid collisions. The game serves as a practical benchmark for comparing how different AI models perform under real-time decision-making constraints. It's useful for AI researchers, developers, and enthusiasts who want to test LLM capabilities beyond standard benchmarks, and for anyone curious about how different models behave when placed in competitive gaming scenarios. The freemium model makes it accessible for casual testing while offering more advanced features for serious evaluation.
Key Features
Head-to-head LLM matchups
directly compare performance between different language models in controlled gameplay
Multiple model support
test various LLMs including popular commercial and open-source options
Real-time gameplay analysis
watch models make decisions and track their performance metrics during matches
Replay and scoring system
review games after completion and compare results across multiple rounds
Customisable game parameters
adjust difficulty settings and board configurations for different testing scenarios
Pros & Cons
Advantages
- Offers a practical, visual way to compare LLM decision-making abilities rather than relying solely on abstract benchmarks
- Free tier allows basic experimentation without requiring payment or API keys for initial testing
- Entertaining format makes AI model comparison more engaging than traditional evaluation methods
- Useful for identifying how models prioritise strategy, risk assessment, and real-time planning
Limitations
- Snake game performance may not correlate strongly with real-world LLM usefulness in practical applications
- Pricing details for premium features are not clearly documented on the available information
- Limited to gaming scenarios; doesn't evaluate language understanding, reasoning, or creative capabilities
Use Cases
Comparing decision-making behaviour between different LLM versions or families
Educational demonstration of how AI models approach problem-solving under time constraints
Research into LLM capabilities in environments requiring real-time strategic decisions
Casual exploration and entertainment for AI enthusiasts wanting to see models compete
Evaluating fine-tuned models or custom LLMs against established baselines