Speaking of China, they are crushing the open source AI model development lately. Last week, Chinese open source models were considered to be roughly 6 months behind the American closed source (oh, the irony) frontier models. This week, Moonshot drops Kimi K3, and closed the gap entirely. This model “delivered performance comparable to OpenAI's GPT-5.5 and ​Anthropic's Claude Opus 4.8, particularly on tests measuring complex, multi-step tasks.” Oh, and it is a fraction of the cost of the American frontier models. This will be interesting to watch play out. In what presents as a thinly veiled attempt to squash smaller competitors, Anthropic has been parading Fable around like it is some sort of technical bio weapon and basically begging the US government to restrict the AI industry. Let’s play it forward and assume the US government agrees and starts restricting access to these powerful models, and China keeps putting out open source models just as powerful, if not more. The ironic but likely scenario would be that if you want freedom in AI land, you flock to the Chinese models over the US frontier labs. This would be hilarious, if not so sad for the state of “freedom” in the United States. When the eventual regulations come to AI models, please remember the principle that this country was built on, freedom of speech. Code is speech.

Read the original article →