DeepSeek says new Flash AI model beats Kimi K3 on cyber, coding benchmarks

DeepSeek says new Flash AI model beats Kimi K3 on cyber, coding benchmarks

DeepSeek V4.1 Flash beats Kimi K3 on coding/cybersecurity benchmarks using 552B MoE with lower inference costs via selective activation. China's efficiency-first AI amid export curbs reshapes foundation model ROI for IT teams evaluating compute-performance trade-offs.