Kimi K3.1 Pro weights appear online

Moonshot AI has published the weights for Kimi K3.1 Pro on Hugging Face. The release drew 198 points and 55 comments on Hacker News. In practical terms, members can now assess an available model against their own workflows instead of relying only on a product description. The model page describes it as an open-weight model that handles text and images, with a context window of about one million tokens.

Moonshot’s page also reports benchmark results across coding, research, office tasks and other areas. Those scores are company-reported, and the page notes that results use particular settings and tools. They’re a reason to identify tasks worth testing, not proof that the model will perform well on your company’s own data or processes. Before putting it into a live workflow, check the outputs against work your team already understands, and consider what handling the model yourself would require.

For an owner or operator, the useful development is the chance to make a direct comparison: take a real task, try the model, and see whether it meets your standards. That can help you decide if it belongs in your AI toolkit, or whether another option fits better. This is the kind of thing we build into an AI command centre, where teams can keep track of models and decide what to test.