DeepSeek's MoE model with 552B total parameters and 8-16B active per token. 1M-token context, MIT license: free commercial use. Built for online serving, too large for local use on a single PC.
Resources
Tools, frameworks, and models for local AI
4 resources found for "MIT"

