Top AI Product

Every day, hundreds of new AI tools launch across Product Hunt, Hacker News, and GitHub. We dig through the noise so you don't have to — surfacing only the ones worth your attention with honest, no-fluff reviews. Explore our latest picks, deep dives, and curated collections to find your next favorite AI tool.


Kimi K3’s 2.8T open weights are live, but running them takes 64 GPUs

Moonshot just put Kimi K3’s weights online, free to download. At 2.8 trillion parameters it’s the largest open-weight model ever shipped. The catch: even after MXFP4 four-bit quantization, the download is 1.4TB.

What you actually get

K3 is a mixture-of-experts model — each token fires only 16 of its 896 experts, about 50 billion active parameters, so it runs cheaper than its size suggests. You also get a 1M-token context window and native multimodal input. But self-hosting means 1.4TB of fast VRAM and at least 64 accelerator cards. Simon Willison put it best: open weights don’t mean you can run it locally.

Weights you’ll still rent

For most teams the real entry point stays Moonshot’s hosted API — $3 per million input tokens, $15 per million output, roughly half the per-task cost of Claude Opus 4.8. The one hard reason to self-host: it’s the only way to keep your data clear of China’s National Intelligence Law. The weights drop the same week as IPO rumors valuing Moonshot at up to $50B.


You Might Also Like


Discover more from Top AI Product

Subscribe to get the latest posts sent to your email.



Leave a comment