Chinese AI start-up DeepSeek has fully open-sourced DeepInfer, an internally developed framework for running AI models more cheaply and efficiently, the South China Morning Post reports. Released on Tuesday, the inference framework is built to support the company's DeepSeek-V4.1-Flash model on both Nvidia's Blackwell graphics processors and Huawei's Ascend 950 chips.
Inference is the process of serving and running a trained model for users, and frameworks like this aim to make deployment faster and less costly. The newspaper said a senior DeepSeek researcher likened the rivalry between open and closed AI models to a fight between light and dark, framing the release as part of the company's push for victory over closed rivals.
Source: South China Morning Post
