DeepSeek serves 150T tokens daily on one cluster, topping four inference providers combined

DeepSeek reports serving around 150 trillion tokens per day on a single cluster, exceeding the combined inference volume of OpenRouter, Fireworks, Together and other major open-model providers, according to an analysis by Nathan Lambert.

2026-10-24 ~ 2026-10-24 · 3 related posts

1 near-duplicate retellings: natolambert