DeepSeek stated it will work with the open-source community to improve inference support and supports discussions for large-scale deployments at the level of 2,000 GPUs.