This repository allows reproduction of Poetiq's record-breaking submission to the ARC-AGI-1 and ARC-AGI-2 benchmarks.

Full analysis is available in our launch post, Traversing the Frontier of Superintelligence.

Our method is now on top of the official leaderboard. More information is in our follow-up post, Poetiq Shatters ARC-AGI-2 State of the Art at Half the Cost.

You can recreate the Gemini 3 points from these charts using this repo.

These are our results on the official leaderboard from ARC Prize, but those problems are kept private.

• Python 3.11+
• API Keys for the models you wish to test (Gemini, OpenAI, etc.)

1. Setup the environment:

   python -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt

2. Create a .env file in the root directory. You must include keys for the models you intend to run.

   GEMINI_API_KEY=...
   OPENAI_API_KEY=...

3. Modify the constants in main.py to set the problem set, number of problems, etc. Then run the script:

   python main.py

4. By default, the code runs the Poetiq 3 config described in the blog post. You can uncomment other ones or modify the config in config.py

If you use this code or these results in your research, please cite our blog post:

Poetiq Team. (2025). Traversing the Frontier of Superintelligence. Poetiq AI. https://poetiq.ai/posts/arcagi_announcement/

For questions or to discuss the future of reasoning, reach out to us at poetiq@poetiq.ai.