DeepSeek has open-sourced DeepInfer, an inference framework for running its AI models at scale. The release gives developers and companies access to software for serving large models and adds another open-source option for AI deployment.
DeepInfer targets AI costs
Inference is the part of AI that runs a trained model and generates an answer. That process can become expensive when users or software agents make large numbers of requests. DeepInfer is designed to make serving DeepSeek models more efficient. TechTarget reported that the framework can accelerate inference and reduce the number of GPUs needed for some workloads.
The software supports DeepSeek-V4.1-Flash. DeepSeek’s public GitHub repository describes DeepInfer as a high-performance, large-scale inference framework and lists it under the Apache 2.0 license.
Two chip platforms supported
DeepInfer supports Nvidia’s SM100 Blackwell graphics processors and Huawei’s Ascend 950 chips, according to the South China Morning Post. That gives the framework a hardware path through both a leading U.S. AI-chip company and a major Chinese chipmaker.
TechTarget reported that the release could help organizations running DeepSeek models on private or on-premises infrastructure. The publication also noted that other open-source inference engines serve different models and hardware.
Open source meets chip rivalry
The release comes as companies try to control the rising cost of AI agents. These systems can make many model requests while completing a task. A more efficient serving layer could reduce operating costs and give developers another option beyond commercial AI APIs.
DeepInfer also arrives amid rising technology tensions between the United States and China. The U.S. Department of Homeland Security ordered DeepSeek products removed from federal civilian information systems and excluded them from related procurements, according to the Federal Register.
References
Lee, C. M. (2026, October 21). DeepSeek seeks “total victory” over closed-source AI with new open-source tool. South China Morning Post. https://www.scmp.com/tech/big-tech/article/3371592/deepseek-open-sources-new-tool-slash-ai-costs-fight-against-closed-source-rivals
Shittu, E. (2026, October 20). DeepSeek makes DeepInfer open source to cut AI costs. TechTarget. https://www.techtarget.com/ai/news/366652363/DeepSeek-makes-DeepInfer-open-source-to-cut-AI-costs
DeepSeek-AI. (2026). DeepInfer: High-performance, large-scale inference framework for DeepSeek models. GitHub. https://github.com/deepseek-ai/DeepInfer
U.S. Department of Homeland Security. (2026, October 20). Notice of order regarding exclusion and removal of DeepSeek products from federal civilian executive branch information technology systems and procurements. Federal Register. https://www.govinfo.gov/content/pkg/FR-2026-10-20/html/2026-21447.htm