[HN Gopher] Kimi k2 largest open source SOTA model?
___________________________________________________________________
Kimi k2 largest open source SOTA model?
Author : ConteMascetti71
Score : 166 points
Date : 2025-07-12 17:26 UTC (5 hours ago)
(HTM) web link (github.com)
(TXT) w3m dump (github.com)
| brcmthrowaway wrote:
| Is Kimi the new deep seek?
| ozgune wrote:
| This is a very impressive general purpose LLM (GPT 4o,
| DeepSeek-V3 family). It's also open source.
|
| I think it hasn't received much attention because the frontier
| shifted to reasoning and multi-modal AI models. In accuracy
| benchmarks, all the top models are reasoning ones:
|
| https://artificialanalysis.ai/
|
| If someone took Kimi k2 and trained a reasoning model with it,
| I'd be curious how that model performs.
| GaggiX wrote:
| >If someone took Kimi k2 and trained a reasoning model with it
|
| I imagine that's what they are going at MoonshotAI right now
| ChrisArchitect wrote:
| [dupe] https://news.ycombinator.com/item?id=44533403
| awestroke wrote:
| This is the model release that made Sam Altman go "Oh wait
| actually we can't release the new open source model this week,
| sorry. Something something security concerns".
|
| Perhaps their open source model release doesn't look so good
| compared to this one
| data_maan wrote:
| "Open source" lol
|
| Open-weight. As usual, you don't get the dataset, training
| scripts, etc.
| mistercheph wrote:
| Wont happen under the current copyright regime, it is
| impossible to train SOTA without copyrighted text, how do you
| propose distributing that?
| irthomasthomas wrote:
| List the titles.
| mixel wrote:
| But probably they don't have the rights to actually train
| on them and that's why they do not publish the list.
| Otherwise it may be laziness who knows
| msk-lywenn wrote:
| Bibtex
| CaptainFever wrote:
| It's not even open-weight. It's weight-available. It uses a
| "modified MIT license": Modified MIT License
| Copyright (c) 2025 Moonshot AI Permission is
| hereby granted, free of charge, to any person obtaining a copy
| of this software and associated documentation files (the
| "Software"), to deal in the Software without
| restriction, including without limitation the rights to
| use, copy, modify, merge, publish, distribute, sublicense,
| and/or sell copies of the Software, and to permit
| persons to whom the Software is furnished to do so,
| subject to the following conditions: The above
| copyright notice and this permission notice shall be included
| in all copies or substantial portions of the Software.
| THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
| EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
| WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
| PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
| AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES
| OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
| TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
| CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
| THE SOFTWARE. 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 shall
| prominently display "Kimi K2" on the user interface of
| such product or service.
| mitthrowaway2 wrote:
| This seems significantly more permissive than GPL. I think
| it's reasonable to consider it open-weight.
| MallocVoidstar wrote:
| 4-clause BSD is considered open source by Debian and the FSF
| and has a similar requirement.
| vessenes wrote:
| I tried Kimi on a few coding problems that Claude was spinning
| on. It's good. It's huge, way too big to be a "local" model -- I
| think you need something like 16 H200s to run it - but it has a
| slightly different vibe than some of the other models. I liked
| it. It would definitely be useful in ensemble use cases at the
| very least.
| summarity wrote:
| 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"
| model.
| refulgentis wrote:
| I write a local LLM client, but sometimes, I hate that local
| models have enough knobs to turn that people can advocate
| they're reasonable in _any_ scenario - in yesterday 's post
| re: Kimi k2, multiple people spoke up that you can "just"
| stream the active expert weights out of 64 GB of RAM, and use
| the lowest GGUF quant, and then you get something that rounds
| to 1 token/s, and that is reasonable for use.
|
| Good on you for not exaggerating.
|
| I am very curious what exactly they see in that, 2-3 people
| hopped in to handwave that you just have it do agent stuff
| overnight and it's well worth it. I can't even begin to
| imagine unless you have a metric **-ton of easily solved
| problems that aren't coding. Even a 90% success rate gets you
| into "useless" territory quick when one step depends on the
| other, and you're running it autonomoously for hours
| segmondy wrote:
| I do deepseek at 5tk/sec at home and I'm happy with it. I
| don't need to do agent stuff to gain from it, I was saving
| to eventually build out enough to run it at 10tk/sec, but
| with kimi k2, plan has changed and the savings continue
| with a goal to run it at 5 tk/sec at home.
| fzzzy wrote:
| I agree, 5 tokens per second is plenty fast for casual
| use.
| refulgentis wrote:
| Cosign for chat, that's my bar for usable on mobile phone
| (and correlates well with avg. reading speed)
| gpm wrote:
| > or even a single socket Epyc system with >1TB of RAM
|
| How many tokens/second would this likely achieve?
| moffkalast wrote:
| Still pretty good, someone with enough resources could distil
| it down to a more manageable size for the rest of us.
| handzhiev wrote:
| I tried it a couple of times in comparison to Claude. Kimi
| wrote much simpler and more readable code than Claude's over-
| engineered solutions. It missed a few minor subtle edge cases
| that Claude took care of though.
| airstrike wrote:
| Claude what? Sonnet? 3.7? 3.5? Opus? 4?
| nathan_compton wrote:
| The first question I gave it (a sort of pretty simple
| recreational math question I asked it to code up for me) and it
| was outrageously wrong. In fairness, and to my surprise,
| OpenAI's model also failed with this task, although with some
| prompting, sort of got it.
| jug wrote:
| I like new, solid non-reasoning models that push the frontier.
| These still have nice use cases (basically anything where logic
| puzzles or STEM subjects don't apply) where you don't want to
| spend cash on reasoning tokens.
| fzysingularity wrote:
| If I had to guess, the OpenAI open-source model got delayed
| because Kimi K2 stole their thunder and beat their numbers.
| irthomasthomas wrote:
| Someone at openai did say it was too big to host at home, so
| you could be right. They will probably be benchmaxxing, right
| now, searching for a few evals they can beat.
| DataDaemon wrote:
| Oops, China is leading with AI, when the Nasdaq investors check
| their AI investments?
| satvikpendem wrote:
| 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 shall
| prominently display "Kimi K2" on the user interface of
| such product or service.
| diggan wrote:
| 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".
| kragen wrote:
| 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 in the most ordinary way, to print or
| display an announcement including an appropriate copyright
| notice and a notice that there is no warranty (or else, saying
| that you provide a warranty) and that users may redistribute
| the program under these conditions, and telling the user how to
| view a copy of this License. (Exception: if the Program itself
| is interactive but does not normally print such an
| announcement, your work based on the Program is not required to
| print an announcement.)_
|
| And the 4-clause BSD license says:
|
| > _3. All advertising materials mentioning features or use of
| this software must display the following acknowledgement: This
| product includes software developed by_ the organization.
|
| Both of these licenses are not just non-controversially open-
| source licenses; they're such central open-source licenses that
| IIRC much of the debate on the adoption of the OSD was centered
| on ensuring that they, or the more difficult Artistic license,
| were not excluded.
|
| It's sort of nonsense to talk about neural networks being "open
| source" or "not open source", because there isn't source code
| that they could be built from. The nearest equivalent would be
| the training materials and training procedure, which isn't
| provided, but running that is not very similar to
| recompilation: it costs millions of dollars and doesn't produce
| the same results every time.
|
| But that's not a question about the _license_.
| moffkalast wrote:
| That's basically less restrictive than OpenStreetMap.
| echelon wrote:
| > This is not open source
|
| OSI purism is deleterious and has led to industry capture.
|
| Non-viral open source is simply a license for hyperscalers to
| take advantage. To co-opt offerings and make hundreds of
| millions without giving anything back.
|
| We need more "fair source" licensing to support sustainable
| engineering that rewards the small ICs rather than mega
| conglomerate corporations with multi-trillion dollar market
| caps. The same companies that are destroying the open web.
|
| This license isn't even that protective of the authors. It just
| asks for credit if you pass a MAU/ARR threshold. They should
| honestly ask for money if you hit those thresholds and should
| blacklist the Mag7 from usage altogether.
|
| The resources put into building this are significant and
| they're giving it to you for free. We should applaud it.
| teiferer wrote:
| > small ICs
|
| The majority of open source code is contributed by companies,
| typically very large corporations. The thought of the open
| source ecosystem being largely carried by lone hobbyist
| contributors in their spare time after work is a myth. There
| are such folks (heck I'm one of them) and they are
| appreciated and important, but their perception far exceeds
| their real role in the open source ecosystem.
| wredcoll wrote:
| I've heard people go back and fortg on this before but you
| seem pretty certain about it, can you share some stats so I
| can see also?
| satvikpendem wrote:
| That's great, nothing wrong with giving away something for
| free, just don't call it open source.
| bhouston wrote:
| Impressive benchmarks!
| emacdona wrote:
| To me, K2 is a mountain and SOTA is "summits on the air". I saw
| that headline and thought "holy crap" :-)
| 38 wrote:
| The web chat has extremely low limits FYI. I ran into the limit
| twice before getting a sane answer and gave up
| exegeist wrote:
| Technical strengths aside, I've been impressed with how non-
| robotic Kimi K2 is. Its personality is closer to Anthropic's
| best: pleasant, sharp, and eloquent. A small victory over botslop
| prose.
___________________________________________________________________
(page generated 2025-07-12 23:00 UTC)