from Hacker News

Seems like GPT non-determinism is mostly from sparse MoE.

Imagine if you use a DeepMind approach of tree traversal for the most promising paths, it can generate deterministic LLM responses.