Proof of concept ✅✅✅
a 100B model running across a handful of gaming GPUs scattered around the world,
at usable speed,
owned by no single party.
afaik this is one of the first few depin experiments proven in the world at this speed and size. most usually solve for one but rarely both.
holy shit, we did it!!
24.77 tok/s on gpt-oss-120b run on four separate 4090's spread across the USA
4 token throughput per traversal lasting 162 ms with speculative decoding