46 comments

[ 4.4 ms ] story [ 78.8 ms ] thread
> 1T total / 32B active MoE model

Is this the largest open-weight model?

This is both the largest oss model release thus far, and the largest Muon training run.
I've only started using Claude, Gemini, etc in the last few months (I guess it comes with age, I'm no longer interested in trying the latest "tech"). I assume those are "non-agentic" models.

From reading articles online, "agentic" means like you have a "virtual" Virtual Assistant with "hands" that can google, open apps, etc, on their own.

Why not use existing "non-agentic" model and "orchestrate" them using LangChain, MCP etc? Why create a new breed of model?

I'm sorry if my questions sound silly. Following AI world is like following JavaScript world.

That's perhaps the best one I've seen yet! For an open weight model, this performance is of course particularly remarkable and impactful.
For what it's worth, I think Kimi's modified MIT license still meets the OSI definition of "open source." For example, the explicitly OSI-approved "Attribute Assurance License"[1] contains similar wording:

> each time the resulting executable program or a program dependent thereon is launched, a prominent display (e.g., splash screen or banner text) of the Author’s attribution information

[1] https://opensource.org/license/attribution-php

It probably doesn't because the attribution requirement discriminates against certain groups (large commercial organisations).
Would be hilarious if Zuck with his billion dollar poaching failed to beat budget Chinese models.
I can't tell if Kimi is quite top tier, but since Llama 4 performed so poorly then yes, this did in fact happen just now.
If the SWE Bench results are to be believed... this looks best in class right now for a local LLM. To be fair, show me the guy who is running this locally...
How does it stack up against the new Grok 4 model?
I really really want to try this model for free since I just don't have a gpu.

Is there any way that I could do so?

Open Router? Or does kimi have their own website? Just curious to really try it out!

Quite impressive benchmark, how come I don't see Kimi in Artificial analysis benchmarks?
How well separated are experts per domain in a model like that? Specifically, if I'm interested in a programming use only, could we possibly strip it to one or two of them? Or should I assume a much wider spread? (And there would be some overlap anyway from the original root model)
The web chat has extremely low limits FYI. I ran into the limit twice before getting a sane answer and gave up
(comment deleted)
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.

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

"Open source" lol

Open-weight. As usual, you don't get the dataset, training scripts, etc.

Open source" lol

It's open-weight. As usual, you don't get the dataset, training scripts, etc.

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