科大讯飞(002230.SZ)：随着国产算力在底层能力上进一步提升，讯飞星火的训练成本还有较大的下降空间

Core Viewpoint - The company has made significant investments in optimizing the training and inference cost efficiency of its large models under limited computing resources, focusing on a fully domestic computing power route rather than relying on NVIDIA cards [1] Group 1: Technological Advancements - Since May 2023, the company has collaborated with Huawei to overcome various technical challenges, including high-speed interconnection networking, hidden computing communication, strong interaction in training and inference, high-throughput inference optimization, and domestic operator optimization [1] - The training efficiency of general large models and deep inference models has improved from an initial 30%-50% to over 85%-95%, benchmarked against NVIDIA's A100 [1] Group 2: Cost Efficiency - As domestic computing power continues to enhance its foundational capabilities, there remains significant potential for further reductions in the training costs of the company's models [1]