LLM Achieves Victory in NetHack: A Milestone
GPT-6 Astra's victory in NetHack demonstrates the gap between game knowledge and execution in LLMs.
GPT-6 Astra has made history by becoming the first LLM agent to successfully complete NetHack 3.6.7. On September 21, after 37,140 game turns, it ascended as a dwarven Valkyrie named CodexDelver. This achievement highlights the gap between game knowledge and execution capabilities in LLMs.
This synthesis was produced from its source by AI; there is no human editor or manual review step. How we work