What's really behind Kimi K3's record numbers

[Redaktion]

China's new flagship model is breaking size records, and the headlines are coming thick and fast. Moonshot itself ranks Kimi K3 behind Claude and GPT. We'll break down what the numbers really mean, whether you'll ever be able to run the model yourself, what it actually costs, and what happens to your input.

https://www.notebookcheck.net/What-s-really-behind-Kimi-K3-s-record-numbers.1346792.0.html

[Eth]

Terabytes

Technically anyone with a 1.5 TB drive can run it. It just will be veeeeery slow (and still much slower on HDD than SSD).

A dedicated switch to turn off training on your content is nowhere in the policy. It points to general data subject rights, which you can exercise through your account settings or by emailing the privacy contact. ChatGPT and Claude, by contrast, put a toggle right in the settings.

But can you guarantee that ChatGPT, Claude and the other proprietary AI makers don't use your data? Also, Kimi K3 is open-weight and this is a huge (not the model, although the model is also huge) and very important fact, even if as a private individuals most are not going to be able to run it.

Your input data will make sure that the next open-weight successor is better.

[Eth]

Also, the power of open-weight models:

Kimi K3 just fixed 15 critical security bugs that Codex and Fable refused because of "cyber guardrails". Hugging Face: We had this experience ourselves this week! Very scary to be guardrailed as a defender when you know attackers are likely bypassing

The remarkable thing about K3 is less the ranking than the pace. A Chinese company is delivering a model on a par with the world's best and is opening up the model data. Two years ago, that would have been hard to imagine.

Exactly!