A-EVO-Lab built an AI system that ran the post-training loop for a 30B model with no human in the loop. Four self-directed rounds on GPU clusters. The autonomously produced model placed 8th of ~4,000 on NVIDIA’s Nemotron Reasoning Challenge — one point behind the top human team.
The same autonomous system has since post-trained the 120B and 550B Nemotron models. This is, to the best of our knowledge, first public evidence at this scale.
A-Evolve is open-source infrastructure that evolves any agent, across any domain, using any evolution algorithm — with zero human intervention. PyPI package a-evolve v0.1.0 released.
Related position paper: “Agentic Evolution is the Path to Evolving LLMs” (arXiv 2602.00359).