DeepSeek's DSpark Boosts LLM Speed by 85%

DeepSeek has released DSpark, an MIT-licensed framework designed to accelerate large language model inference by up to 85% without altering model outputs. The system uses a speculative approach where a smaller scout model predicts likely text sequences ahead of time, allowing the main model to verify and accept correct guesses rapidly. The release comes amid rising geopolitical tensions over AI, with the U.S. government moving to restrict models from Anthropic and OpenAI. DeepSeek's continued open source contributions signal its growing influence on global AI development.