Boring AI News

Axion 1.0: Open Agent Foundation Models, Simplified

Published 8 November 2026

Axion 1.0 is an open agent model trained with two supervised fine-tuning stages and reinforcement learning. It learns reasoning, tool use, and long interactions through automatically verified tasks, with all data methods, prompts, and code released for reproduction.

papers
Read the original on arxiv.org →

More from 8 November 2026

Every Monday we pick the five stories from the week worth keeping.

Read the latest AI Digest