DeepSeek Says Enterprises With 2,000 GPUs Can Contact It Directly

DeepSeek said in its V4.1 Flash open-source announcement that companies with 2,000 GPUs and a storage cluster can contact it directly. According to ChainCatcher, V4.1 Flash remains open source under the MIT License and does not add licensing conditions based on company revenue or business scale. Companies with sufficient resources can deploy the model themselves, while direct contact with DeepSeek appears aimed at ultra-large-scale technical cooperation. DeepSeek has not announced an official private deployment product or paid service. The company said that after previous open-source model releases, deployment adaptation was mainly handled by inference frameworks and hardware vendors such as SGLang and TensorRT-LLM. This time, DeepSeek said it is proactively leaving direct contact information for large-scale deployments and will continue supporting the open-source community in adapting new models for inference.