Kimi K3.1 Pro
Kimi K3.1 Pro is an advanced open-weight multimodal reasoning model developed by Moonshot AI as the flagship successor to Kimi K3. Built as a 3-trillion-parameter-class Mixture-of-Experts (MoE) architecture, it features 2.8 trillion total parameters with approximately 104 billion active parameters per token, paired with a massive 1-million-token context window. Designed for complex agentic workflows, long-horizon coding, and deep reasoning, the model incorporates extensive post-training compute scaling and large-scale joint reinforcement learning across multiple downstream domains.
The model exhibits strong performance across multi-step execution, tool-assisted problem solving, and specialized technical domains such as parametric 3D modeling and financial research. By optimizing its long-context inference and training infrastructure, Kimi K3.1 Pro delivers high accuracy on software engineering benchmarks like Terminal-Bench 4.0. It is accompanied by a companion variant, Kimi K3.1 Pro Highspeed, designed to offer significantly faster throughput for interactive use cases.
Moonshot AI provides access to Kimi K3.1 Pro through its API platform, development environments like Kimi Code, and enterprise deployments featuring data privacy controls. Supporting native text, image, and video understanding, the model represents a major advancement in open intelligence systems for autonomous agents and complex knowledge work.
Create with Crafiq
Generate images, 3D models, video and audio in one studio.
Explore the studioHow Kimi K3.1 Pro ranks
Kimi K3.1 Pro is highlighted in the table below. Switch the metric to see how the ordering changes.