Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model
1–10 of 194 posts
Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model
#2> 1T total / 32B active MoE model
Is this the largest open-weight model?
Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model
#3Big release - https://huggingface.co/moonshotai/Kimi-K2-Instruct model weights are 958.52 GB
Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model
#4> 1T total / 32B active MoE model Is this the largest open-weight model?
I believe so.
Grok-1 is 341B, DeepSeek-v3 is 671B, and recent new open weights models are around 70B~300B.
Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model
#5Big release - https://huggingface.co/moonshotai/Kimi-K2-Instruct model weights are 958.52 GB
Paired with programming tools like Claude Code, it could be a low-cost/open-source replacement for Sonnet
Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model
#6This is both the largest oss model release thus far, and the largest Muon training run.
Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model
#7Big release - https://huggingface.co/moonshotai/Kimi-K2-Instruct model weights are 958.52 GB
Paired with programming tools like Claude Code, it could be a low-cost/open-source replacement for Sonnet
According to the bench its closer to Opus, but I venture primarily for English and Chinese.