Imagine if China actually do care about dangerous frontier AI being available in other countries, and Beijing allows K3 because they know K3 (knowledge cutoff early 2026) isn't actually the most advanced models in China.
Artificial Intelligence thread
- Thread starter 9dashline
- Start date
More than these models, I think China fear a world of haves and haves not. Where a group tech feudal lords in Silicon Valley controls who can get access AI technology and who doesn't and If AI live the hype that means extreme inequality worldwide.
And they lived that scenario themselves, the US tech war against China was a US elite war to gatekeep China from accessing AI tech.
Also undermine the US companies monopoly on AI.
So there is probably more specific trained advanced models in China, that the public will never have access.
I mean, if you just take one step back and think about it, without prior expectations its should actually be quite implausible that Chinese labs publish their latest techniques or open-weight their latest models on day one.
Beijing also isn't exactly known for being laissez faire when it comes to security.
IMO the real "security" issue with frontier models is probably detection of deliberate vulnerabilities and backdoors rather than creation, US fears frontier models for the same reason they fear Huawei: it makes it much harder to spy on everyone through vulnerabilities, which also means it's in China's interest to make it as widely available as possible
Most public AI models are generalists, so my guess is that they have models as smart or even smarter than Kimi K3 but heavily trained for military and cybersecurity applications.
But given the US paranoia about Chinese tech they could release those ones too, I have my doubts that the stooges will run ANY Chinese AI models in their networks.
I'm definitely seeing some rankings where K3 is finishing even ahead of Fable. I'm not saying it is better and it probably isn't, but it's still a big deal for K3 to be this competitive.
I'm putting it through a whole host of research topic right now
More or less equal. So 6 months behind is officially 2-3 months behind. Officially.
K3 is a 2.9T parameters sparse model that cost $15 per 1M output token or far less on subscription.
Fable is a 6T parameter (or 10T depending on how you count) dense model that cost $50 per 1M output tokens and can only be accessed via API.
You have to completely ignore Anthropic being an order of magnitude behind in the singular most important metric in performance to say Kimi is 2-3 month behind. The better question is how long will it take for Anthropic to make make a Fable that's actually usable in term of cost and efficiency.
K3 basically just destroyed the last bastion of American AI where they try to charge 100x for marginal capability delta, now they have to try charge 100x for no capability delta, or for regular users, for all practical purposes, 10x for an inferior product in Opus.
To put it another way, I can right now on Kimi subscription go to down and use K3 for everything without a single thought on token usage, if you do that with Fable you'd go broke in an instant. The real metric is, between two companies both with $100 budget, how much can each build if one use Fable and one uses K3. It's not even a competition.
Last edited: