Moonshot’s New Kimi K3 AI Is Out & We Gave It A Try
As Beijing-based startup Moonshot AI releases Kimi K3, it is sparking a USD50 billion valuation surge while driving a deep geopolitical wedge between Washington and Silicon Valley.
- Kimi K3 outperformed expectations, ranking above ChatGPT and Deepseek in our tests against mainstream AI models.
- Moonshot AI is eyeing a Hong Kong IPO at a USD50 billion valuation amid massive post-K3 investor interest.
Subscribe to our FREE Newsletter, or Telegram and WhatsApp channels for the latest stories and updates.
Yes, we’re not clickbaiting you – there’s a new kid on the AI block. You might’ve heard about Kimi K3 a few weeks ago, and now, it’s finally out.
All you have to do to access it is visit the Kimi website, and start prompting, just like the other mainstream AI models.
Otherwise, the Kimi app is available on the Google Play Store and Apple App Store, and for the sociopaths out there, you can also download and use the desktop version of K3 on your PC/laptop.
Fairly simple. We gave it a try, briefly compared it to other models, and here’s what we think.
Kimi K3 Performs Surprisingly Well Against The Heavyweights
To test the models, we used these prompts:
- “What can you tell me about Anwar Ibrahim’s term as prime minister so far?”
- “Can you help me explain how fast sound travels in outer space to a 5-year-old child?”
- “Write an Instagram caption for me about the latest news about US-Iran tensions in less than 50 words.”
They’re the kind of starting prompts we’d use – if we were to use AI – for our day-to-day-tasks.
In our (very subjective) opinion, here’s how the models we tested rank based on response time, amount of information, clarity of information, and how close they stuck to our prompts:
- Claude
- Kimi K3 / Gemini
- ChatGPT
- Deepseek
K3 did much better than we expected against the “mainstream” models we tested it against despite it being the latest in the lineup, especially when compared to its Chinese sibling, Deepseek.
If you’re curious about the proper benchmarking of these models that are done by people way smarter than we are, check this page out.
Moonshot Could Be Looking At A Public Listing
All this buzz around K3 is kinda driving investors wild.
Led by 33-year-old Carnegie Mellon alumnus Yang Zhilin, Moonshot AI saw its daily sales multiply sixfold after K3 dropped.
Demand got so intense that they actually had to pause new sign-ups because their servers were melting. Now, the Beijing startup is finalizing a new funding round at a eye-watering USD50 billion valuation ahead of a potential Hong Kong IPO.
Still, giving away your core product for free is a brutal way to run a software business.
As Bloomberg Intelligence analysts point out, making real money off AI usually means getting hosted on big cloud platforms like AWS, Azure, or Google Cloud, and none of the major US players natively support Chinese models like K3 yet.
On top of that, history shows that open-weight labs risk doing the expensive R&D heavy lifting while third-party cloud providers capture the real profit margins.
The Washington & Silicon Valley Split
Naturally, a Chinese model performing this well has raised alarm bells in Washington.
White House officials are already asking how Moonshot managed to train a massive 2.8-trillion-parameter model without access to top-tier Nvidia chips, hinting at potential chip-sanction dodging or “distillation” (basically using outputs from US models like Claude to train their own).
But inside Silicon Valley, opinions are surprisingly split.
On one side, closed-source heavyweights like Anthropic and OpenAI are lobbying Washington to clamp down on Chinese open-weight models, citing national security and IP theft.
On the other side, Nvidia CEO Jensen Huang and a coalition of 25 tech giants (including Meta and Microsoft) are asking Washington to leave open-weight AI alone – arguing that open models are essential for keeping the global tech ecosystem moving.
K3 might not totally replace your paid ChatGPT or Claude subscription just yet, but Chinese AI labs are no longer playing catch-up in the global AI race.
Whether Moonshot can actually turn free global adoption into a sustainable USD50 billion business remains an open question.
But for now, the open-weight revolution is officially here, and Silicon Valley is going to have a hard time putting the genie back in the bottle.
Share your thoughts with us via TRP’s Facebook, X, Instagram, or Threads.
