DeepSeek open sources DeepInfer, the engine behind V4.1 Flash, for Nvidia and Huawei chips
DeepSeek has published DeepInfer, the in house software that serves its V4.1 Flash model in production, as open source under the Apache 2.0 license. It runs on Nvidia SM100 GPUs and Huawei Ascend 950 chips, ships with DeepSeek's own benchmarks and leaves out the parts needed for a complete serving platform.