DeepSeek open-sources DeepInfer inference framework, discloses cluster throughput and inference revenue figures

[AINews] DeepSeek DeepInfer: Open Source SOTA Inference running 155T toks/day/cluster making $22m/day at ~80% margin for a 81 day payback

DeepSeek released DeepInfer, the inference framework behind V4.1 Flash, along with the deepseek-recipe API handling components. Some KV-cache management and load-balancing infrastructure is not fully open yet. The company also published production data, including a 96% input cache hit rate.

Why it matters: DeepSeek's production numbers, including the 96% input cache hit rate, give needed context for its throughput and revenue estimates.