Dumb-1.1 RC1 - a new SoTA 19M llm!
This LLM is stupid, but it provides interesting English and performance per parameter of a slightly higher model, and tells a more beautiful lie.
In addition, for the first time in the Dumb series, it is an AI that has a correct answer rate of more than 1/4 with ARC-Easy.
- Downloads last month
- 74
Safetensors
Model size
18.9M params
Tensor type
F32
·