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

Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

twitter.com

Post reply on HN
Print view View original

141–150 of 194 posts

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#141

Post by user_7832 »

mindcrime wrote: post #104
kragen wrote: post #78

Earlier quoted context omitted.

I feel like those restrictions don't violate the OSD (or the FSF's Free Software Definition, or Debian's); there are similar restrictions in the GPLv2, the GPLv3, the 4-clause BSD license, and so on. They just don't have user or revenue thresholds. The GPLv2, for example, says: > c) If the modified program normally reads commands interactively when run, you must cause it, when started running for such interactive use…

It may not violate the OSD, but I would still argue that this license is a Bad Idea. Not because what they're trying to do is inherently bad in any way, but simply because it's yet another new, unknown, not-fully-understood license to deal with. The fact that we're having this conversation illustrating that very fact. My personal feeling is that almost every project (I'll hedge a little because life is complicated) s…

I'm of the personal opinion that it's quite reasonable for the creators to want attribution in case you manage to build a "successful product" off their work. The fact that it's a new or different license is a much smaller thing.

A lot of open source, copyleft things already have attribution clauses. You're allowed commerical use of someone else's work already, regardless of scale. Attribution is a very benign ask.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#142

Post by cristoperb »

sagarpatil wrote: post #134

All the AI models are no using em-dashes. ChatGPT keeps using them even after explicitly told not to. Anybody know what’s up with these models?

I don't know, but as someone who likes using em-dashes in my writing it is disappointing that they have become a marker of LLM slop.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#143

Post by sophiebits »

ensignavenger wrote: post #137
kragen wrote: post #78

Earlier quoted context omitted.

I feel like those restrictions don't violate the OSD (or the FSF's Free Software Definition, or Debian's); there are similar restrictions in the GPLv2, the GPLv3, the 4-clause BSD license, and so on. They just don't have user or revenue thresholds. The GPLv2, for example, says: > c) If the modified program normally reads commands interactively when run, you must cause it, when started running for such interactive use…

The OSD does not allow for discrimination: "The license must not discriminate against any person or group of persons." "The license must not restrict anyone from making use of the program in a specific field of endeavor. For example, it may not restrict the program from being used in a business, or from being used for genetic research." By having a clause that discriminates based on revenue, it cannot be Open Source.…

This license with the custom clause seems equivalent to dual-licensing the product under the following licenses combined:

* Small companies may use it without attribution

* Anyone may use it with attribution

The first may not be OSI compatible, but if the second license is then it’s fair to call the offering open weights, in the same way that dual-licensing software under GPL and a commercial license is a type of open source.

Presumably the restriction on discrimination relates to license terms which grant _no_ valid open source license to some group of people.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#144

Post by londons_explore »

numpad0 wrote: post #139
overfeed wrote: post #121

Earlier quoted context omitted.

Also works perfectly fine in fire-and-forget, non-interactive agentic workflows. My dream scenario is that I create a bunch of kanban tickets and assign them to one or more AI personas[1], and wake up to some Pull Requests the next morning. I'd me more concerned about tickets-per-day, and not tk/s as I have no interest in watching the inner-workings of the model. 1. Some more creative than others, with slightly diffe…

> I create a bunch of kanban tickets and assign them to one or more AI personas[1], Yeah that. Why can't we just `find ./tasks/ | grep \.md$ | xargs llm`. Can't we just write up a government proposal style document, have LLM recursively down into sub-sub-projects and back up until the original proposal document can be translated into a completion report. Constantly correcting a humongous LLM with infinite context len…

In my experience, this sort of thing nearly works... But never quite works well enough and errors and misunderstandings build at every stage and the output is garbage.

Maybe with bigger models it'll work well.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#145

Post by winter_blue »

wongarsu wrote: post #130
summarity wrote: post #64

Earlier quoted context omitted.

Reasonable speeds are possible with 4bit quants on 2 512GB Mac Studios (MLX TB4 Ring - see https://x.com/awnihannun/status/1943723599971443134 ) or even a single socket Epyc system with >1TB of RAM (about the same real world memory throughput as the M Ultra). So $20k-ish to play with it. For real-world speeds though yeah, you'd need serious hardware. This is more of a "deploy your own stamp" model, less a "local" mod…

Reasonable speeds are possible if you pay someone else to run it. Right now both NovitaAI and Parasail are running it, both available through Openrouter and both promising not to store any data. I'm sure the other big model hosters will follow if there's demand. I may not be able to reasonably run it myself, but at least I can choose who I trust to run it and can have inference pricing determined by a competitive mar…

I’m actually finding Claude 4 Sonnet’s thinking model to be too slow to meet my needs. It literally takes several minutes per query on Cursor.

So running it locally is the exact opposite of what I’m looking for.

Rather, I’m willing to pay more, to have it be run on a faster than normal cloud inference machine.

Anthropic is already too slow.

Since this model is open source, maybe someone could offer it at a “premium” pay per use price, where the response rate / inference is done a lot faster, with more resources thrown at it.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#146

Post by randomNumber7 »

satvikpendem wrote: post #75

This is not open source, they have a "modified MIT license" where they have other restrictions on users over a certain threshold. Our only modification part is that, if the Software (or any derivative works thereof) is used for any of your commercial products or services that have more than 100 million monthly active users, or more than 20 million US dollars (or equivalent in other currencies) in monthly revenue, you…

This is just so Google doesn't build a woke version of it and calls it gemini-3.0-pro

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#147

Post by cubefox »

fzysingularity wrote: post #69

If I had to guess, the OpenAI open-source model got delayed because Kimi K2 stole their thunder and beat their numbers.

According to the benchmarks, Kimi K2 beats GPT-4.1 in many ways. So to "compete", OpenAI would have to release the GPT-4.1 weights, or a similar model. Which, I guess, they likely won't do.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#148

Post by mrob »

diggan wrote: post #77
satvikpendem wrote: post #75

This is not open source, they have a "modified MIT license" where they have other restrictions on users over a certain threshold. Our only modification part is that, if the Software (or any derivative works thereof) is used for any of your commercial products or services that have more than 100 million monthly active users, or more than 20 million US dollars (or equivalent in other currencies) in monthly revenue, you…

That seems like a combination of Llama's "prominently display “Built with Llama”" and "greater than 700 million monthly active users" terms but put into one and masquerading as "slightly changed MIT".

The difference is it doesn't include Llama's usage restrictions that disqualify it from being an Open Source license.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#150

Post by mindcrime »

user_7832 wrote: post #141
mindcrime wrote: post #104

Earlier quoted context omitted.

It may not violate the OSD, but I would still argue that this license is a Bad Idea. Not because what they're trying to do is inherently bad in any way, but simply because it's yet another new, unknown, not-fully-understood license to deal with. The fact that we're having this conversation illustrating that very fact. My personal feeling is that almost every project (I'll hedge a little because life is complicated) s…

I'm of the personal opinion that it's quite reasonable for the creators to want attribution in case you manage to build a "successful product" off their work. The fact that it's a new or different license is a much smaller thing. A lot of open source, copyleft things already have attribution clauses. You're allowed commerical use of someone else's work already, regardless of scale. Attribution is a very benign ask.

I personally have no (or at least little) problem with attribution. As you say, quite a few licenses have some degree of attribution required. There's even a whole dedicated (and OSI approved) license who's raison d'être is about attribution:

https://en.wikipedia.org/wiki/Common_Public_Attribution_Lice...

What I'm saying, if I'm saying anything at all, is that it might have been better to pick one of these existing licenses that has some attribution requirement, rather than adding to the license proliferation problem.

Post reply on HN