China-based Moonshot AI released Kimi K3 yesterday. The release had been widely hyped up, largely by early testers sharing impressive 3D demos and outputs from the model. But the real surprise came with the benchmarks — Kimi K3 scored 57 on the Artificial Analysis Intelligence Index, ahead of every AI model on earth except Fable 5 and GPT 5.6 Sol, and was in touching distance of those two. In addition, it was priced at roughly half of these models. And Moonshot says that it’ll open-source the weights in a few days, effectively making this intelligence available to anyone for free who has the hardware to run them.
This made the broader AI community sit up and take notice of what just happened. “kimi k3 is a big moment with multiple implications for the entire industry,” wrote Sriram Krishnan, who until recently was working on AI regulation with the US government.
Researchers from US-based AI labs also said that the new Chinese model changed things. “the era of the chinese labs being far behind is over, Kimi is at least on par with the modern public frontier models. people have to think differently now without any competitive margin built in,” said roon, who is an anonymous OpenAI researcher.
Pytorch co-creator Soumith Chintala, who is currently at Thinking Machines, called it a world-class model.
Investor Jason Calacanis talked about how the model would cause a general acceleration in all kinds of capabilities. “It’s happening folks… things accelerated more in the last 30 days from a dozen players than in the last year Open Source models are compounding Frontier Models are refining When open source hits robotics, self driving and life sciences things are going to get wild,” he said on X.
South Park Commons’ Aditya Agarwal said that he was immediately switch off Anthropic’s Fable. “I am literally switching models off of Fable right now for our systems. This isn’t trying to be hyperbolic….but why would you pay the price if there is a good and free alternative?” he said.
Some felt that the release laid to rest claims by US companies that Chinese labs were distilling their models. “US labs gonna end up distilling Chinese models,” said Stability AI founder Emad Mostaque.
The sentiment was echoed by the co-CEO of Arena. On Arena’s public scorecard, Kimi K3 had debuted at top place, ahead of Fable 5 and GPT 5.6 Sol.
Some wondered about the implications of the release for the AI race between the US and China. “There is absolutely no question that the “Chinese models” are economic weapons (purposefully or not) that threaten hundreds of billions of US investment. Paradoxically, those weapons are so efficient thanks to two key positive forces: free market and open source,” wrote Meta’s Francois Fleuret.
Huggingface CEO Clement Delangue said that countries that were open-sourcing their models had a structural advantage. “I’ve said it many times: the countries or companies that are leading in open science and open source AI will start leading the frontier a few years later as it accelerates AI progress massively! That’s how the US took the lead,” he said.
Wharton professor Ethan Mollick felt that the release would ease US regulatorty pressure on its own models.
And popular model jailbreaker Pliny the Liberator said that this was the flippening. “the Open-Source Capabilities Flippening cometh,” he posted.
It really does seem like it’s a bit of a flippening. The US created nearly all the technologies that required for the AI revolution to take place, including the invention of GPUs, the Image Net research, the invention of the Transformer, and the release of ChatGPT in late 2022. But just four years later, China appears to be able to build AI models that rival those built by the top labs in the US. These models are also cheaper and open-source, which means that anyone with the weights can use them. It remains to be seen how the user reaction to Kimi K3 is, but based on benchmarks alone, its release, perhaps even more than DeepSeek, seems like an epochal moment for the AI race.