Caiyue Star and Shanghai Jiao Tong University released Alpha-R1, an 8B parameter financial reasoning model. Utilizing semantic gate technology and two-stage reinforcement learning, it dynamically matches strategies to market states. Out-of-sample tests on 2025 data show Alpha-R1 significantly outperforms general LLMs and traditional methods, achieving superior annualized returns and cross-environment generalization. The open-source model offers transparent, traceable decision-making for intelligent investment research and risk monitoring.