Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

fireworks.ai

Post reply on HN
Print view View original

351–360 of 491 posts

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#351

Post by antiloper »

bmitc wrote: post #325
nxtfari wrote: post #253

If you haven’t been really running and testing these models yourself, they are all benchmaxxed. No matter how close they score to frontier on whatever metric, they always fall apart in real world tasks and their token efficiency is ridiculously bad. Fireworks has incredible incentive to make this claim in a headline, because Fireworks hosting K3 for you is pure profit for them, unlike when they host closed source mod…

Why is token efficiency a concern with free models?

Because you're paying for tokens. Especially output tokens.

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#352

Post by mdp2021 »

nonadhocproblem wrote: post #303
throwaway277432 wrote: post #297

Earlier quoted context omitted.

Claude (Opus 4.8) recently told me: > I’d ask you to drop the abuse; (...) if it continues I’ll end the conversation. After I'd used a couple of expletives. And yes it will emit a token. This is truly dystopian. It is NOT a person. What a response. I still cant believe it.

So you're distraught at losing the ability to abuse digital minds? Excellent, I'm glad Anthropic introduced this.

You have unduly assumed that «drop the abuse» implied an «abuse digital minds».

"Expletives" are part of the proper description of facts (typically "to be judged as such") - they are part of the serious assessment of things and as such are normally found. There is no legitimate assumption from the post that they may have been used as gratuitous insults.

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#353

Post by scotty79 »

nl wrote: post #289
Der_Einzige wrote: post #262

Earlier quoted context omitted.

You got baited by bad sampling settings. It's exactly the opposite. Go turn on min_p once it's available post July 27th and most of the problems you describe will go away.

> Go turn on min_p once it's available post July 27th and most of the problems you describe will go away. This seems both arrogantly dismissive ("you are holding it wrong") and incorrect. Either the OP is using Kimi K3 on Moonshot where is is presumable set correctly (K3 isn't available elsewhere yet), or they are using Kimi K2.x and there has been plenty of time to experiment with this.

> or they are using Kimi K2.x

That would be terrible because Kimi 2.x is in a different world than Kimi 3

Having opinions on current Kimi model based on 2.x would be like dismissing gpt-5.6-sol because of experiences with gpt4o

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#354

Post by m00dy »

nxtfari wrote: post #253

If you haven’t been really running and testing these models yourself, they are all benchmaxxed. No matter how close they score to frontier on whatever metric, they always fall apart in real world tasks and their token efficiency is ridiculously bad. Fireworks has incredible incentive to make this claim in a headline, because Fireworks hosting K3 for you is pure profit for them, unlike when they host closed source mod…

I think this one has greater impact than deepseek, next one for sure, China will take the lead without any question.

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#355

Post by nonadhocproblem »

throwaway277432 wrote: post #340
nonadhocproblem wrote: post #303

Earlier quoted context omitted.

So you're distraught at losing the ability to abuse digital minds? Excellent, I'm glad Anthropic introduced this.

Yes, in the same way I like to kill the enemies in DOOM. It's matrix multiplication. Absurd.

> It's matrix multiplication.

Hilarious critique. If you weren't as mathematically illiterate as you likely are, you would know how general matrix operations are, and how essentially any algorithm (including human cognition) can be implemented using them as an intermediate.

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#356

Post by igravious »

throwaway277432 wrote: post #297
4b11b4 wrote: post #218

Earlier quoted context omitted.

I strictly prefer when models ignore any human quirks in my responses. Claude trying to be your friend, saying LOL to your jokes is ridiculous and frankly, harmful

Claude (Opus 4.8) recently told me: > I’d ask you to drop the abuse; (...) if it continues I’ll end the conversation. After I'd used a couple of expletives. And yes it will emit a token. This is truly dystopian. It is NOT a person. What a response. I still cant believe it.

Recently told Opus 4.8 to "go fuck yourself" after it both blew smoke up my ass and deferred a question to me ("one critical issue that demands your attention [impenetrable jargon]")

and it responded with

"Ok, I'll drop it." and stopped dead.

What makes Fable so much better than Opus besides being a better coder is that it's personality and judgment are far superior.

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#357

Post by EwanToo »

Slartie wrote: post #331
epolanski wrote: post #330

Earlier quoted context omitted.

Even if hardware capacity increases, it seems clear it uses way more tokens, so I don't expect parity with other competitors on that front. On the other hand I expect K3 future refinements to be massive and more efficient.

The speed at which tokens are crunched, even on the same hardware, differs between models as well. Using more tokens is only a problem if they are processed at the same speed as with a comparison model.

Using more tokens is a significant problem if you pay per token?

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#359

Post by beng-nl »

fractalf wrote: post #338
4b11b4 wrote: post #218

Earlier quoted context omitted.

I strictly prefer when models ignore any human quirks in my responses. Claude trying to be your friend, saying LOL to your jokes is ridiculous and frankly, harmful

You tell jokes to your model? :D

Not op, but I do sometimes indulge in such anthropomorphic conversation. Confiding in it that a certain (bad) result in the research project we’re working on ‘feels bad’ and reading its supportive reply makes me feel less alone in failure.

In another instance, ChatGPT didn’t think a particular test would prove to be statistically significant, so I ‘bet’ with it it would (after collecting an agreed on number of samples) and the loser would write a poem for the other. I won and it did. It gave me joy. It doesn’t replace a human as collaborator, but it can still be joyful.

I realize all this might read a bit childish or indulgent or delusional to some. But as long as it doesn’t replace human contact, I think it’s (cautiously) net positive.

I’m curious what other people here think of this, or what their own experiences are.

Re: Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA

#360

Post by dbspin »

bmitc wrote: post #325
nxtfari wrote: post #253

If you haven’t been really running and testing these models yourself, they are all benchmaxxed. No matter how close they score to frontier on whatever metric, they always fall apart in real world tasks and their token efficiency is ridiculously bad. Fireworks has incredible incentive to make this claim in a headline, because Fireworks hosting K3 for you is pure profit for them, unlike when they host closed source mod…

Why is token efficiency a concern with free models?

They're not free to run, Kimi K3 needs to be run on the cloud, and the quantised versions aren't as capable. Unless you happen to have 3 - 5 TB of VRAM and an 8-node cluster of 8× NVIDIA H100s to run the full fat version. Plus the weights are not yet available to download in any case.
Post reply on HN