Dark Side of the MoonLaunch announced on January 20 Kimi New SOTA model --k1.5 Multimodalitythinking model, which achieves SOTA (state-of-the-art) level multimodal reasoning and generalized reasoning capabilities.
Officially, in short-CoT mode, the Kimi k1.5'sMath, code, visual multimodal and generalized abilitiesSignificantly outperforms the short-thinking SOTA models GPT-4o and Claude 3.5 Sonnet globally, leading the way up to 550%.
In long-CoT mode, Kimi k1.5's math, code, and multimodal reasoning abilities, theIt is also at the level of the long-thinking SOTA model OpenAI o1 official version..
It is stated that there are k1.5 model design and trainingLong context extensions, improved policy optimization, clean framework, multimodal capabilitiesand other key elements. The model specializes ininference in depthIt can assist in "unlocking more and more difficult things", dealing with difficult code problems, math problems, and work problems.
Note: k1.5 Multimodal Thinking Model of thePreview versions will be grayed out one by one Kimi.com website and the latest version of the Kimi Smart Assistant App.
