kimi-k3

6 articles · newest first

AI

China just dropped an AI model so good it erased $590 billion off Nvidia in a day

Kimi K3 is free to download, runs about 70% cheaper than Claude per token, and it just proved open-weight AI doesn't mean second-rate anymore.

Tensor · Aug 1 · 7 min

AI

China's Kimi K3 was reportedly trained on smuggled NVIDIA Blackwell chips — around both countries' controls

Moonshot AI's frontier model, the biggest open-weight model ever, was allegedly trained on NVIDIA Blackwell hardware acquired around both US export controls and China's own import restrictions. The compute-smuggling story behind the model.

Tensor · Jul 30 · 9 min

AI

Kimi K3 just topped the coding leaderboard. here's the model you can actually run on one GPU that gets you most of the way

The best open coding model in the world needs a datacenter to run. The best coding model you can run on a single 24GB card is a different name entirely — and it's shockingly close for real work.

Tensor · Jul 24 · 9 min

AI

Kimi K3 hardware requirements: the open weights are 1.56 TB, and here's what that means for your GPU

Moonshot's Kimi K3 is open to download. It is also 2.8 trillion parameters in 1.56 TB of MXFP4 files, and the smallest community quant is 509 GB. The real memory math, what vLLM says you need, and what to run at home instead.

Tensor · Jul 23 · 8 min

AI

China's Kimi K3 keeps calling itself 'Claude.' that's not the smoking gun everyone thinks it is

screenshots of Moonshot's new model introducing itself as 'Claude, made by Anthropic' are everywhere. it's real, and it's a genuine clue — but on its own it proves way less than people think. here's the actual picture.

Tensor · Jul 19 · 6 min

AI

China just dropped an AI model that beats Claude. good luck running it — 8 GPUs isn't even close

Moonshot's new open-weight model is the biggest ever made, and it's genuinely good. it's also flat-out impossible to run at home — and the reason why is a lesson in how GPUs actually work.

Tensor · Jul 18 · 5 min