[chvid]

Hard to be fully open source if you train on copyrighted material.

Anyway. Deepseek is the most open of the sota models.

[MoonGhost]

Did they open their datasets already? Would be nice to have 'thinking' part.