Loading video...
Video Failed to Load
Kimi K3: China just beat Claude AI at coding (full video below)
29,162 views • 2 months ago •via X (Twitter)
24 Comments

Kimi K3: China just beat Claude AI at coding

at frontend. not "coding". and it thinks its claude.

Great to witness Chinese LLM achieving competitive coding capability against top international models. Personally, I’m more concerned about its performance on theoretical mathematical AI modeling tasks.😊

Turning a hand drawn character into a playable fantasy world is a really fun way to show off the coding chops

Dude, it's frontend.

Moonshot is the Amodeo's worst nightmare. They are so good.

You might be able to "download" it... but to run it locally you'll need eye-watering amounts of compute 🫰🖥️

Okay... so, it's all about coding, right? How about 99% of other activities that AI was supposed to do instead of humans? Can Kim do that now? Just thinking...🤔

Cant even solve this

Definitely cheap models are the way to go although Kimi3 ain’t that cheap. $15/m output tokens is still a lot. Def cheaper than the thieves at misanthropic though

nope. yet

The trust question is always 1st, regardless of the position question. President Trump points out in his national address on July, 16, 2026, that the CCP paid American influencers, media networks, educational systems, politicians, and local US governments through reward programs not to say anything bad about them. Regulations do not score high in trust with the CCP.

It will be interesting to see what claude comes up with

China gets one moment and it’s all over! 😂 when your a influencer you must ride the wave

That's cool, but I'm curious: does it handle architecture, governance, and orchestration at scale? Because those seem to be where most tools hit the wall in enterprise. What's your take?

You had it right the first time brother! 😜 🦁 🚀 Let’s go. 🦁

This is very interesting!!!🤨

先别急着封神,得看测的是啥题吧

I think the fear of sharing data / IP with chinese labs will forever be a limitation. Cheaper tokens cant fix trust issues

Wow, that's impressive! It's great to see advancements like this happening.

Unknown: @thuztra just beat all the traditional traders💯

Let me know if you want to be a free beta tester for my AI futures trade call system. For ES NQ RTY and CL.

Remember, @techleadhd has a coin solana:5wNia6Yx6dfsFia2e8eEZE4vZ4qnFbgQf75dirWhpump

Frontier CUDA kernel optimization still requires Fable. Kimi K3 couldn't handle the tasks we threw at it. Yet for nearly everything else, Kimi K3 has become our default. The implication is difficult to ignore: Anthropic's long-term sustainability looks increasingly uncertain.

