Latent Space·· 2 hours agoFeaturedAI Score80
DeepSeek open-sources DeepInfer inference stack, reporting 155T tokens/day per cluster
[AINews] DeepSeek DeepInfer: Open Source SOTA Inference running 155T toks/day/cluster making $22m/day at ~80% margin for a 81 day payback
AI summary
DeepSeek open-sourced DeepInfer, the inference engine behind V4.1 Flash, and published production figures including 155 trillion tokens/day from one cluster.
Why it matters
The production numbers come with the cache-hit and cost assumptions that decide whether the margin holds.
Source: Latent Space · latent.space