A team from China's Tsinghua University has released Chitu, an open-source AI framework challenging NVIDIA's market leadership. Chitu supports various language models and can operate on Chinese-made chips, potentially reducing reliance on NVIDIA GPUs. Developed by startup Qingcheng.AI, led by Professor Zhai Jidong, Chitu claims a 315% increase in inference speed and 50% reduction in GPU usage compared to foreign frameworks. The project, backed by Beijing's municipal AI fund, aims to provide versatility across different hardware environments.