DeepSeek releases DeepInfer source code for running AI models
The framework runs DeepSeek’s production services and is available under the Apache 2.0 licence. It currently supports only DeepSeek-V4.1-Flash. According to MarkTechPost, the release targets NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs, but ships with offline inference only rather than the full capabilities used for serving customers.
- According to The Next Web, DeepSeek published the code on GitHub.
Covered by 8 publishers within 15 hours of the first report.
- SiliconANGLE: DeepSeek open-sources DeepInfer, the inference framework behind its production services
- Latest South China Morning Post: DeepSeek seeks ‘total victory’ over closed-source AI with new open-source tool
Reporting8
The Next Web DeepSeek open-sources DeepInfer, the engine that runs its AI models · Ana Maria Constantin
MarkTechPost DeepSeek Open-Sources DeepInfer, the Inference Framework Behind DeepSeek-V4.1-Flash · Asif Razzaq
The New Stack DeepSeek open-sources the inference engine behind its API. And Huawei gets equal billing with Nvidia. · Paul Sawers
Gizmodo DeepSeek Dials Up the Pressure on OpenAI and Anthropic With Open-Source Inference Tool · Webb Wright
PYMNTS DeepSeek Open Sources Inference Tech to Lower AI Costs · PYMNTS
SiliconANGLE DeepSeek open-sources DeepInfer, the inference framework behind its production services · Duncan Riley
The Information DeepSeek Open-Sources Framework Behind Its High-Margin AI Inference · Juro Osawa
South China Morning Post DeepSeek seeks ‘total victory’ over closed-source AI with new open-source tool · Chong Ming Lee