moonshotai/Kimi-K3.1-Pro: Open-weight multimodal agentic model with 2.8T parameters

huggingface.co7 min read

Moonshot AI has released Kimi K3.1 Pro, an open-weight, natively multimodal agentic model featuring a 1-million-token context window and a Mixture-of-Experts architecture with 2.8 trillion total parameters. Post-trained using large-scale joint reinforcement learning across hundreds of domains, the model delivers notable improvements in coding, financial research, factual accuracy, and 3D creation. It employs Gated DeepSeek Sparse Attention layers for efficient long-context modeling and is accessible through the company's API platform. Open article →

More on this topic