DeepSeek open-sources inference optimizations with 60–85% faster generation [pdf]

DeepSeek open-sourced DSpark, a set of inference optimizations that achieve 60–85% faster generation on large language models, releasing the paper and code on GitHub.

Uh oh There was an error while loading. Please reload this page . deepseek-ai / DeepSpec Public Notifications You must be signed in to change notification settings Fork 24 Star 368 Expand file tree / DSpark paper.pdf Copy path More file actions More file actions Latest commit History History History 706 KB / DSpark paper.pdf Copy path File metadata and controls 706 KB Download raw file Edit and raw actions