Anthropic Ran an AI Across 15 Smart TV Boxes to See If It Could Spread

Anthropic’s Frontier Red Team bought 15 off-brand Android TV boxes, roughly $ 55 each with 4GB of RAM, wired them together, and split a 26-billion-parameter model across them. Each box held a couple of the model’s layers. The fleet produced 0.39 tokens per second. That number anchors the research published on 12 October by William Zhou and Daniel Freeman. For scale, the paper cites a third-party benchmark in which a single H100 serves the same model at about 1,950 tokens per second, though that figure is measured at high concurrency with speculative decoding rather than in a like-for-like test against…

First reported by How2Shout. This page is a short excerpt; the full text is theirs.