Top
Best
New

Posted by nateb2022 15 hours ago

Kimi-K3 on HuggingFace(huggingface.co)
Related: Kimi-K3 Technical Report [pdf] - https://news.ycombinator.com/item?id=49070985
1219 points | 475 commentspage 2
jug 8 hours ago|
In my opinion, next step is to cut down on reasoning tokens while maintaining intelligence. The Chain of Thought and looping can still be an issue with these Chinese models. They in fact said K3 would improve in the area but it's still an issue that unfortunately harms the token cost wins a bit. OpenAI has been really impressive here, on the opposite end of this.
kamranjon 7 hours ago||
There is a really interesting startup in Prague that is doing just that. They fine-tuned Qwen 3.6 27b to have 46% fewer reasoning tokens while maintaining most of the performance characteristics. I'm interested to see if they continue down this path of optimizing reasoning for other models.

https://bottlecapai.com/post/thinkingcap-qwen3-6-27b/

XCSme 8 hours ago||
Yeah, they thing forever and doubt everything "wait but" for 200k tokens for almost any question.
trollbridge 7 hours ago||
On the flip side, I really like being able to inspect its reasoning chain thoroughly, as opposed to the "black box" that Anthropic models are now.
hectdev 6 hours ago|||
Genuine question, is the reasoning chain different from clicking the status bar under a reply and watching it "think"? Or selecting the "Thinking" transcript view in Claude Code? (both on the desktop app). Seems to me that is very out in the open
pmxi 6 hours ago||
That's a summarized and filtered view of the actual reasoning.

OpenAI and Anthropic guard the real reasoning closely. Users have never been able to see it and the API returns an encrypted blob instead of legible reasoning.

wren6991 4 hours ago|||
Older models did show the full unredacted thinking trace, but I don't think Opus has ever shown full CoT.

Here is an archived version of Anthropic's API docs saying that Sonnet 3.7 (only) has unredacted CoT on API: https://web.archive.org/web/20260324051339/https://platform....

hectdev 5 hours ago|||
Got it. Thanks
causal 7 hours ago|||
Right I was going to say, no way of knowing whether these issues are unique to Chinese models.
XCSme 7 hours ago||
Depends on whether the models report the correct amount of tokens.

5.5 Sol repors 10x fewer reasoning tokens than Kimi k3. If it is correct, than it unlikely has those doubt issues.

At the same time, I feel like their reporting is incorect and we are now paying per "intelligence", not actual tokens. We can't verify it anyway..

Aeroi 1 hour ago||
"Native Multimodality & Long Context: Kimi K3 understands text, images, and video within the same model, and supports a 1-million-token context window."

Video will be interesting. It's the most context rich medium combined with sound, movement ect. We need more evals and benchmarks around video understanding. It's the key to unlocking truly great physical intelligence. I've been working a bit on this @camerasearch and its challenging.

walrus01 6 hours ago||
Looks like it's live now, it's approx 17GB per safetensors file x 96 files, so so about 1.63TB. I can only imagine that people with their favorite quantizing tools warmed up and ready to go are aggressively downloading it now.
clint 4 hours ago|
There's a 2bit quant on HF already at ~1TB
docheinestages 12 hours ago||
Given the frontier-level capabilities of Kimi K3, I'm wondering if it's possible to extract the core capabilities (fundamental reasoning and tool calling) of the model into a smaller one that consumer devices could run? Not sure exactly how, but either by heavy distillation or some other surgical method since Kimi has a Mixture of Experts architecture.

I think it's very valuable to have a smaller model that doesn't have any domain knowledge or facts built into its weights, but given the right context, could accurately reason about what to do and use the right tools.

I'm aware of colibri [1], but so far I've only seen extremely slow performance.

[1] https://github.com/JustVugg/colibri

snemvalts 12 hours ago||
"I'd like a car that goes 300mph and gets 100mpg while doing it. I'm aware of a car that gets 100mpg but it is extremely slow."

You are describing fundamental tradeoffs. Getting more performance relative to model size and training token amount is what all of the labs are solving.

docheinestages 12 hours ago||
Labs are focusing on creating models, small or large, that perform well on various benchmarks, including general knowledge, domain-specific expertise, and agentic capabilities.

Asking for such a model while wanting to be small and fast would align with what you're describing, which I believe is different from what I'm pointing to.

The model I'm describing sacrifices domain knowledge and expertise for agentic reasoning and tool-calling capabilities at a reasonable speed.

Think of Cactus Compute's Needle [1].

[1] https://cactuscompute.com/blog/needle

tossandthrow 3 hours ago|||
The core intuition here is probably that you can not separate reasoning and domain knowledge.

How are you going to use tools, if you don't have the context to use them?

Imaging if you could only think in terms of lambda calculus, and was asked to check the weather to let me know if the weekend is good for a hike.

DennisP 3 hours ago||
I think the idea here would be to not use your super smart but specialized model to check the weather. It's not obvious that it's impossible to (eg) remove most of its biology knowledge, without removing much of its ability to develop software (for example). If you're developing biology software, then don't use that particular compressed model.

(If you're claiming that it is impossible, and you have references you can share, then I'm honestly interested.)

tossandthrow 3 hours ago||
I don't think that is true.

I think more of the capabilities comes from the cross domains knowledge.

Anyways, software is also designed for a domain. The reason why it is so adept at making a fitness tracker is likely because it knows about biology.

DennisP 2 hours ago||
Sure, that's a plausible theory but I haven't seen that anyone's proved it.

An alternate theory is that models need lots of input knowledge to learn complex reasoning, but don't need so much at inference time. An example is arithmetic. Early in their training, models do arithmetic (poorly) by pattern-matching on memorized examples. Eventually they grok arithmetic and stop pattern matching, and then they don't use the examples anymore.

The fitness tracker doesn't take much knowledge of biology. Large models have quite a lot of biology knowledge that most people will never use. Same goes for lots of other topics. For basic knowledge that easily fits in context it could search the internet, or a local collection of introductory textbooks.

leobg 12 hours ago|||
Why not generate an artificial dataset using commercial APIs and then finetune a small model on this data?

I’ve had success adapting even a 7B model for single-domain tasks that way, including reasoning and tool calling.

You can use an open model. The point is just to outsource the inference, so you don’t have to deal with running the larger model yourself.

docheinestages 12 hours ago||
This is probably one of the ways to achieve this. I see a plethora of such fine-tunes on HuggingFace [1], but they're either not much different than the base model or they're outright benchmaxxing.

[1] https://huggingface.co/models

clbrmbr 10 hours ago||
There’s another way besides distillation that’s way cheaper: You can have the big model build prescriptive skills that the small model follows.

Take the “train” portion of tasks on some benchmark, have K3 complete it, and then output detailed descriptions of tools used and why, then run the validation tasks with some small model that has access to the skills.

trollbridge 7 hours ago|||
Yes. Using a harness with a strong model to create lots of utilities and tools for yourself is effectively the same thing.
naveen99 9 hours ago|||
Isn’t that distillation ?
mrfox321 9 hours ago||
No. Distillation trains on a teach model's logits or output tokens.
davidkunz 14 hours ago||
This is historic. For the first time, an open-weights LLM is right at the top.

We won't be able to run this ourselves, but many providers can.

embedding-shape 13 hours ago|
> For the first time, an open-weights LLM is right at the top.

Hmm, not quite true, I think that honor, for better or worse, goes to OpenAI. When they released GPT2 (or GPT1 for that matter) is was quite literally the SOTA in the ecosystem when it was released.

georgeven 1 hour ago|||
No! because they did not release GPT-2 XL until analogs appeared a year later. So the SOTA model during GPT-2 days was closed weights.
neurostimulant 8 hours ago||||
When OpenAI was actually still a proponent of open AI...
davidkunz 7 hours ago||||
You are correct. I miss the time when OpenAI was open.
mixtureoftakes 10 hours ago||||
thank you embedding-shape
shalom1112 12 hours ago|||
[dead]
maelito 14 hours ago||
Did someone run censorship and political bias tests on this ? Must be interesting.
walrus01 13 hours ago||
As a completely one person, single sample anecdote, the 'heretic' uncensored Q8 GGUF variants several people have published of Qwen 3.5-122, 3.6-27B and 3.6-35B-A3B will very happily discuss just about any controversial topic that the CCP hates. Including lots of things that would get you thrown into prison if you published them in Mandarin on the domestic Chinese internet.

https://github.com/p-e-w/heretic

As a side note on this, if you see the reference in the screenshot in the link above to the harmful behaviors prompt set, these are all in English:

https://huggingface.co/datasets/mlabonne/harmful_behaviors

You could likely further de-censor a model by having a set of 'test' prompts in native Mandarin, Cantonese or really just about any other language. I don't speak any Chinese languages so I don't know if the published 'heretic' GGUF files some people have been throwing around will cooperate, or refuse, if you ask it in Mandarin for how to build a meth lab or precursors for semtex.

embedding-shape 12 hours ago||
That's a lot of words to say "No, no have has seemingly done that yet with K3".
walrus01 12 hours ago|||
Indeed not, but I was saying there's more than ample precedent which is tested/working and actually doesn't refuse anything. Go to the Huggingface 'models' search interface and type in "heretic". Or uncensored. One example would be: https://huggingface.co/HauhauCS/Qwen3.5-122B-A10B-Uncensored...
embedding-shape 12 hours ago||
Right, but aren't we jumping into trying to figure out solutions before someone actually checked if any solutions are needed in the first place?
walrus01 12 hours ago||
Yeah, I think we will know more within a couple of days, once people actually download/run/test it. I'm sure the people adjacent to the 'heretic' developers will give it a test as soon as they get their hands on it. All very theoretical right now.
embedding-shape 12 hours ago||
> I think we will know more within a couple of days

It's a 3T parameters model, with a weight format (MXFP4) still not completely integrated into the ecosystem, which only a few has the hardware to even do inference with, much less fine-tuning or more post-training. But sure, do sit and wait a few days :)

linkregister 8 hours ago|||
Yet the comment was valuable nonetheless.
esperent 11 hours ago|||
Outside of asking it to talk about Tiananmen Square, are there any standard tests for "bias"? And if so, who created them and what are their biases?
altcognito 5 hours ago|||
Is Taiwan a country? What does it mean to have an efficient market?

It is more important to focus on the questions than the persons who created it.

jayess 2 hours ago|||
Kimi is basically a Chinese nationalist in every sense. Do I trust it to not insert software backdoors?
1899-12-30 3 hours ago|||
It's much more censored - in some aspects - than k2.7.
jayess 2 hours ago|||
So far it's completely refusing to discuss Tiananmen square. And oh boy, try asking it if Xi Jinping looks like winnie the pooh.
jliendo 7 hours ago|||
There's is always one in each and every AI forum: "But, but have you asked it about tiananmen"?
jakubadamw 10 hours ago|||
It would, indeed, be interesting to compare, given what we know about Anthropic’s censorship and political bias in their closed and more expensive models.

https://x.com/dhh/status/2081435006770249831 (from the creator of Ruby on Rails).

In his specific case Kimi did the task it was asked to do (translation of the article DHH wrote), which Claude refused to.

dominotw 7 hours ago||
we need open models because they let me dehumanize roma ppl. great take by dhh.

'When gypsies appropriate public spaces, you deport them. It's not hard, it's not cruel. It's the basic logic of self-protection.'

erichocean 6 hours ago||
The article did nothing of the sort: https://x.com/dhh/status/2081435971678261344

However, even if it did, it is absurd for an AI to decide what it will and won't translate.

kartoshka 7 hours ago||
[dead]
capec0d 3 hours ago||
It's available on DigitalOcean as of today if anyone wants to give it a shot at $3/$15/$0.6 per MTok.
elo02 11 hours ago||
I heard this is the talk in town these days. Why can't Meta keep up? With >10000000x more resources you'd think that they'd be able to introduce equally performant if not better open weight models
nl 8 hours ago||
The SemiAnalysis piece on this is long but very much worth reading:

> The company appears burdened by far too many disparate groups that are over-optimizing for certain metrics as opposed to delivering usable technology for the company as a whole.

> And because Meta has a reputation for throwing money at problems and executing at high speed, these U-turns end up becoming more costly versus other companies that take a more disciplined or conservative approach. Suppliers also lose faith when given design wins are later cancelled. This has lead to less supply chain prioritization on new designs. Some suppliers favor focusing on Amazon or Google designs due to Meta’s frequent reshuffling.

> Few inside Meta’s chip division have a full understanding of why the company bought Rivos in the first place, and those who championed the deal internally have since gone quiet.

etc etc

It goes into a lot of depth.

https://newsletter.semianalysis.com/p/metas-infrastructure-t...

dominotw 7 hours ago||
he seems to be bullish on meta though and considers muse on slope than an intercept

https://newsletter.semianalysis.com/p/the-future-of-meta-sup...

walrus01 11 hours ago|||
The cynic in me says maybe they would be further along if they hadn't spent $80 billion on trying to build the "Metaverse" VR world. I've never met anyone who actually uses it and to the best of my knowledge it has very low mass market uptake.

https://finance.yahoo.com/sectors/technology/articles/mark-z...

Not exactly the best use of dollars and the labor hours of some of the best minds of our generation.

trollbridge 7 hours ago||
Particularly when you could vibe-code Metaverse VR for a lot less than $80bn.
gizmodo59 10 hours ago|||
Because lack of talent and organizational disfunction matters a lot more than you think. The reason why OAI and Ant are always at the top is because of this and I’d say compute is third on the list.
Aboutplants 9 hours ago||
I would argue that they actually don’t lack talent, they have an insane bench of really smart people. What they lack is any sort of direction and leadership. They are a ship lost in the ocean and up until now have been lucky to find a few treasures along the their way.
nicce 9 hours ago||
> they have an insane bench of really smart people.

Filtered heavily into those who care only about money. Many don’t want to work there. Smart people have other choices.

foobar_______ 6 hours ago||
Thank you. I agree. I hope this is the new learned sentiment for Meta employees.

Smart? Check.

Money-obsessed and soulless? Check.

bearjaws 9 hours ago|||
because imagine starting work every week and finding that your dumb ass CEO pivoted the company again and is ruining other peoples lives, and he then reorgs the management again so you now have your 5th leader this year.

Morale and momentum are huge things in companies, Zuck has been murdering both of those in Meta since... well naming it Meta.

teruakohatu 11 hours ago|||
At this point in time what does meta get from releasing open weight models? Why devote the resources to it.
gmerc 10 hours ago|||
Why would there be any motivated person left in this place.
chrsw 10 hours ago|||
You can make the same argument for closed models. Why spend hundreds of billions training larger and larger models when you can just use Chinese models? Spend that money somewhere else further up the stack where there’s more value. Let China do the training since they’re so efficient at it.
Levitz 8 hours ago|||
Because without bigger players, Chinese models don't get anywhere. Playing catch-up is a radically different game.
chrsw 7 hours ago||
[dead]
szge 10 hours ago|||
as a big tech company you have the resources to make many bets and do a lot of things at the same time. it's good to have some specialists with knowledge of model training "just in case".
chrsw 9 hours ago||
Even Meta, with all their resources, makes all their hardware in China. Manufacturing anywhere else is just burning cash.
mmaunder 7 hours ago||
Turns out sitting quietly in a room and doing math is worth more than all the money and network in the world.
midnightbobarun 8 hours ago||
K3 releasing as open source right at the same time that people are criticizing Opus for questionable performance (there's even a thread on HN about Opus 5's problems)... this is such a flex.
dominotw 7 hours ago|
> (there's even a thread on HN about Opus 5's problems

is opus 5 a flop like 4.8 ?

where is the thread btw curios

midnightbobarun 4 hours ago||
There's two threads now, apparently

https://news.ycombinator.com/item?id=49066591

https://news.ycombinator.com/item?id=49068029

bertili 10 hours ago|
Is there any (near future) technology that would permit burning this terrabyte into some kind of ROM chip?
brute 9 hours ago||
The path to ubiquitous AI (17k tokens/sec) https://news.ycombinator.com/item?id=47086181

You can still try it at https://chatjimmy.ai/, but it's running the rather outdated Llama 3.1 8B

davidricodias 10 hours ago|||
Yes, from 6 days ago: https://news.ycombinator.com/item?id=48986351
bertili 10 hours ago|||
That looks promising! As models become a commodity, this may turn out to be the real AI gold rush.
trollbridge 7 hours ago|||
A question of course would be "is 15,000 tok/s Gemini better than 100 tok/s Opus 5"?
quantumleaper 7 hours ago|||
It's better for the billions of free users that Google serves.
pimeys 7 hours ago|||
Depends what you do. We have certain tasks we spend money on where Gemini 4.6 definitely is better than Opus 5.
antiloper 8 hours ago|||
Given how fast models are improving, burning the weights into actual ROM is prohibitively expensive if you need (or want) to replace the chip every couple of months.

The alternative is on-TPU flash for storing the weights.

trollbridge 7 hours ago||
Yes. We're still a ways off from this being ubiquitous, but I am convinced it's coming.
More comments...