LG’s recent Exaone release is a pretty awesome local model for the size. A little deep-fried and repetitive, but great for code and stuff, which is especially notable since most decent models tend to only be good at Mandarin Chinese and English.
That is not the precedent. Many 32B class (aka 16GB-24GB VRAM GPU) models are Apache licensed. Hence, the ML tinkerer community has pretty much forgotten about it.
LG’s recent Exaone release is a pretty awesome local model for the size. A little deep-fried and repetitive, but great for code and stuff, which is especially notable since most decent models tend to only be good at Mandarin Chinese and English.
…Except they slapped an insane license on it. Basically you sign away your life away even looking at it: https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B/blob/main/LICENSE
That is not the precedent. Many 32B class (aka 16GB-24GB VRAM GPU) models are Apache licensed. Hence, the ML tinkerer community has pretty much forgotten about it.