Good Start Labs - AI Model Training & Evaluation 〜 With Games
Train and evaluate AI models with games
Simulations that measure and improve your models' key capabilities: Long-horizon reasoning, tool use, enterprise functionality, and more.
Get in touch
Trusted by
TRAIN
Improve models with reinforcement learning environments
- Models learn generalizable skills from millions of interactions
- Skills learned transfer to real world agent tasks
- Variations reward distinct skills and personality traits
Key Partner
Partnered with Arcee to support the training, evaluation, and data sourcing efforts of the Trinity model family.
LIVE FEEDBACK
Make your AI better by having it learn from real people
- Players provide actionable feedback on how to improve models
- Games are continuous, scalable sources of quality data
- Enables real world performance insights
Our Bad Cards agent runs 20,000+ games weekly, getting live human feedback.
Bad Cards' 2 million+ users easily add AI agents to games with their friends.
EVALUATE
Measure what traditional benchmarks miss
Tested across thousands of simulated scenarios covering long-term reasoning, tool use, negotiation, theory of mind, and honesty to better understand the personality of a model.
We evaluated GPT-5 ahead of release to inform model decisions. We identified high steerability and prompt sensitivity, enabling OpenAI to provide instructions on how to prompt GPT-5 and a prompt optimizer.
LEADERBOARDS
See where top AI models rank in our games
200,000+
Matches played
Public
Open, transparent leaderboards
DiplomacyHumor
Last updated: 2026-1-15Last updated: 2026-01-15
Overall Performance
How quickly and how often does it win?
Rank| Model| Organization| Score
Loading...
Published research
Advancing the field of AI training through games
Opening the Door: Democratizing Diplomacy
Accepted to NeurIPS workshop. We created the first Diplomacy environment where even small models can play full games. Our goal was to give people tools to understand how different AI models make decisions.
AI Diplomacy: Towards Understanding AI
We launched our first project AI Diplomacy. Different models revealed their true character: some betrayed without hesitation, while others, like Claude, chose principles over victory. We made AI behavior visible through Diplomacy, for 45,000 unique Twitch viewers, many experiencing AI for the first time.
Interested in collaborating?
Get in touch
Ready to improve your models?
Talk to us about using our game environments for evaluation, training data, or RL environments.
We typically respond within 48 hours
Name *
Email *
Company
Website
What are you reaching out about? *Select an optionImproving AI modelsGetting AI live in gamesCollaborating on researchOther
Anything else?
Send Message