LLMs · via X
DeepSeek open-sources DeepInfer, its production inference stack
DeepSeek released the framework that serves DeepSeek V4.1 Flash at scale. Anyone running open models can now use the same high-throughput, low-latency stack to cut compute costs.