Original caption
DeepSeek just released an inference architecture upgrade that makes compatible AI models up to 400% faster! It’s called DSpark, and it’s already making the recent DeepSeek models more efficient and much cheaper to use. This is showing how the new focus in the AI race is about inference speed. DeepSeek is showing us how to compete the open source way! #breakingnews #deepseek #aitechnology #ainews #gigaqian