The Chinese retailer JD.com also joins the ranks of model trainers, releasing a 48B-A3B model trained on 20T tokens. They claim it can rival Qwen and GLM-4.7 Flash, which is an impressive feat for their first release!
Specs
Params48B (3B activated)
LicenseModified MIT
Tags
modelsmodels/llmsmodels/llms/reasoningmultilingualartifacts/19
Related Models



