After DeepSeek R1, there's new model from China that outperforms Claude Sonnet 3.5 & GPT-4o. Kimi is a multimodal model with advanced Chain-of-thoughts reasoning and real-time web search.
After DeepSeek R1, there's new model from China that outperforms Claude Sonnet 3.5 & GPT-4o. Kimi is a multimodal model with advanced Chain-of-thoughts reasoning and real-time web search.
Contribute to MoonshotAI/Kimi-k1.5 development by creating an account on GitHub.
You're viewing a single thread.
How does it compare to r1 though. They seem to have avoided comparing them
From my anecdotal experience they're pretty similar. A couple of nice things with Kimi is that it can do internet searches, and you can paste images in for it to analyze.
Internet searches don't depend on the model, you just need the infrastructure to feed it in
Sure, I'm just noting the feature their service currently provides that DeepSeek doesn't.
Deepseek does support search on web and mobile.
Oh haha, I didn't realize that. I just kind of assumed it worked with the existing corpus of data. With Kimi it explicitly tells you when it troves websites and how many sites it looked at.