DeepSeek Open-Sources DeepInfer Production Stack to Cut AI Inference Costs

Chinese AI developer DeepSeek released DeepInfer as open-source software on Tuesday, sharing the internal production stack it uses to run artificial intelligence models at scale and reduce inference compute costs. The software currently supports DeepSeek's V4.1-Flash model, with the Hangzhou-based company stating that support for future AI systems will follow. DeepInfer uses techniques including distributed inference to spread AI workloads across hardware, aiming to increase model output speed and extract more performance from chips, though no independent benchmarking of its efficiency gains has been conducted yet. The release allows cloud hosting providers and enterprises to deploy DeepSeek's cost-cutting methods on their own infrastructure, while external developers can study and adapt the open-source code for their own inference systems. "We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable." — DeepSeek