Decentralized AI training networks like @Pluralis' Agora are the only true counterbalance to centralized AI. https://x.com/Pluralis/status/2065899910080115117

Decentralized AI training networks like @Pluralis' Agora are the only true counterbalance to centralized AI. https://x.com/Pluralis/status/2065899910080115117

“there are no H100, B200's” https://x.com/pluralis/status/2065899910080115117

“there are no H100, B200's” https://x.com/pluralis/status/2065899910080115117

A year ago it was impossible to train LLMs on consumer GPUs. Now it’s happening in real time and the parameter count is going up. https://x.com/pluralis/status/2065899910080115117

Pluralis is training a model on consumer GPUs with the explicit goal of enabling you to own a piece without waiting for a trillion$ IPO, and also becoming the best (open source) LLM thanks to distributed compute. https://x.com/pluralis/status/2065899910080115117

Pluralis is training a model on consumer GPUs with the explicit goal of enabling you to own a piece without waiting for a trillion$ IPO, and also becoming the best (open source) LLM thanks to distributed compute. https://x.com/pluralis/status/2065899910080115117

I'm not here to convince you, but the state of the art of decentralized training on DiLoCo style approaches alone is closer to 100-200B parameters, which is commercially viable size. (See latest release from @MacrocosmosAI, e.g.) @Pluralis is at 8B with *model-parallel* training that breaks up the