Motif Launches Open-Weight Motif-3-Beta AI Model With 13 Billion Active Parameters and 256K Token Context
Korean lab Motif has published the open-weight Motif-3-Beta model on Hugging Face, featuring a 314 billion-parameter MoE architecture with 13 billion parameters active per token. The model runs a 256K token context window and activates eight of its 384 available experts during inference.
Performance comparisons cited in release materials place the model on par with larger foundations such as MiniMax M3 and DeepSeek V4 Pro. The company also distributed a technical report and kernel code with the weights, detailing a proprietary activation function and a modified attention mechanism.
From the sources (4 posts)
@eliebakouchhuge open weight release, motif (korean company) just released a 13B active 314B total MoE performing on par with bigger models like minimax M3 and deepseek v4 Pro they incorporate their own research bets with per expert activation functio
@eliebakouchi like this lab a lot, they also always release tech report and also kernels so very exicting! model
@_akhaliqMotif-3-Beta just dropped on Hugging Face ~314B total parameters / ~13B active per token (sparse MoE) 256K context length (262,144 tokens), natively long-context Sparse routing: 384 experts with 8 activated per token, plus 1 shared exper
@huggingfaceRT @eliebakouch: huge open weight release, motif (korean company) just released a 13B active 314B total MoE performing on par with bigger m…