HUGE announcement in the AI world.

I knew this day was coming, but I didn’t think it would happen this soon.

I’ve been a big advocate for open-weight AI models. If you have not heard of an open-weight AI model, these are models you can download completely for free, and then run, study, and modify the model on your own local infrastructure.

While there are some US based AI model providers like Google, who have released open-weight AI models like Gemma, the most powerful models in the US have come from companies like OpenAI and Anthropic, but their models (ChatGPT and Claude) are completed closed, with no ability to download and run yourself.

China has been leading the world with their open-weight models, but these models have always received the same critique that they aren’t as good as ChatGPT or Claude.

Until this past week…..

Moonshot AI just released Kimi K3, and the benchmarks (do your own testing too) are pretty outstanding for an open-weight model.

Based on an aggregation of 9 popular benchmarks, Kimi K3 is only beat out by Claude Fable 5 and GPT-5.6.

Keep in mind, this model is MASSIVE, with 2.8 trillion parameters, so running on your local machine is not very realistic, but if it performs even close to Fable 5 or GPT-5.6, it seems like the US AI models might be in trouble.

What do you think, has China caught up in the AI race? Does Kimi K3 provide strong competition?