DeepSeek releases DeepInfer open-source infrastructure and V4.1 Flash model

Chinese AI company DeepSeek has released DeepInfer, an open-source software infrastructure designed to run AI models at lower costs. The system currently supports the DeepSeek-V4.1-Flash model and is compatible with both Nvidia Blackwell GPUs and Huawei Ascend 950 chips. By utilizing distributed inference techniques, DeepInfer aims to optimize resource consumption, reduce energy use, and make advanced AI more accessible by maximizing performance from existing hardware.

Additionally, DeepSeek has launched an upgraded version of its open AI model, DeepSeek-V4.1-Flash-1017. This minor update improves reasoning, coding, agent tasks, and image understanding while offering faster task completion and shorter responses. The model weights are released under the permissive MIT license, allowing for commercial use, modification, and distribution. In internal comparisons, the model showed competitive results in coding and automation benchmarks against major American models.