lucataco/
deepseek-v3-64k:latest

18 7 months ago

A strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.

... /
params
1c16f08d3b69 · 164B
{
"num_ctx": 64000,
"stop": [
"<｜begin▁of▁sentence｜>",
"<｜end▁of▁sentence｜>",
"<｜User｜>",
"<｜Assistant｜>"
]
}