Distributed AI inference and training has one major disadvantage over centralized AI: it runs over the internet.
These models are massive, requiring huge amounts of high speed data transfer to train, and to efficiently serve inference.
While @DoubleZero tacked blockchain and market data first, distributed AI is very much something we're building for.
If you're building distributed AI and latency or data movement is an issue, let's talk.
1/ We published our first technical report today.
We ran a 229B model split across five consumer GPUs in five countries over the public internet and measured 12.6 tok/s interactive, 194 tok/s batched.
With cryptographic receipts on every request.
doi.org/10.5281/zenodo…
