China's Moonshot AI released the weights of Kimi K3.1 Pro, its flagship multimodal model, on Hugging Face, so organizations can run it themselves. Until now the model was available only through Kimi's apps and API.
The model uses a mixture-of-experts architecture with 2.8 trillion total parameters, about 104 billion of them active per inference, and a context window of around one million tokens. According to TechNode, the release builds on Kimi K3 with expanded reinforcement learning for coding, 3D creation, financial research and long-running agent tasks.
Huawei announced inference support through vLLM Ascend and SGLang on its Atlas A3 and Ascend 950 systems. The license terms for the weights were not detailed in the report.