aichannel

My biggest complaint about Kimi K2 is that it’s slow. They’re fixing it. And yes, more open source from China

We've heard your feedback — Kimi K2 is SLOOOOOOOOOOOOW 😭
Especially for agentic apps, output tokens per second really matters.

The main issue is the flooding traffic and huge size of the model, we are actively working on inference optimization and BUY MORE MACHINES!

Speed

Kimi.ai

My biggest complaint about Kimi K2 is that it’s slow. They’re fixing it. And yes, more open source from China

https://x.com/kimi_moonshot/status/1945064408809660743?s=46&t=iLSaYO2O7t6yPdmEXeqMdA

Applied AI, databases, data, vibe coding (sic.), open source, community/devrel, remote. 

itinerant businessperson. globalist.

Kuala Lumpur, Federal Territory of Kuala Lumpur, Malaysia

Hi @ccarella.eth I am still waiting for capitalism (tm) producing good open source models! Sorry but this reminded me of you 🫢

Still waiting for communism to produce a good model.

Not long ago, I'd have said they produced DeepSeek, which Perplexity managed to post-train, and give us DeepSeek that ran without resource constraints. But a lot has changed since then. The best model from China to me, so far, is Qwen - I would give it a go if you haven't.

I just want to use the best model in the world. I don’t care where kr comes from. For me that’s Claude Opus rn.

I have actually settled on Claude Sonnet 4 :)
(OK, I lie, I also use o3 and gemini pro 2.5)

They're all a good mix. And yeah, origin matters least to me - just get the job done, as best as you can, and as fast as you can

do you have sufficient compute for local execution? i'm still very much a laptop + ssh kinda guy, so i'm ok with it not being local per se (but it could be nice!)