Orion-16B trained across multiple continents on consumer and datacentre GPUs over ordinary internet, roughly the equivalent of 18 B200s…

Macrocosmos@MacrocosmosAI

Orion-16B trained across multiple continents on consumer and datacentre GPUs over ordinary internet, roughly the equivalent of 18 B200s reached from the long tail. Machines dropped, throughput swung, and the run kept training without anyone stepping in.

Open on X ↗

Taosis indexes announcements, articles, posts and releases about Bittensor subnets from their own channels and keeps a record of each. The text above is the source’s; the figures are read from the chain by Taosis.