Latent Space·· 5 h ago
[AINews] DeepSeek DeepInfer: Open Source SOTA Inference running 155T toks/day/cluster making $22m/day at ~80% margin for a 81 day payback
[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 has open-sourced DeepInfer, the production inference engine behind V4.1 Flash, and released the companion deepseek-recipe protocol conversion library, with support for NVIDIA SM100 and Huawei Ascend 950.
Selection record
Threshold 76Media and individualsFirst 82Second 88
AdmittedSum of both 170 ≥ twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed：DeepSeek开源推理框架及生产数据
- Why it was chosen
- Open-sources a production inference stack and discloses operating data including cache hit rates and per-card revenue, showing how high cache hit rates underpin inference economics.
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: Latent Space · latent.space