Chat with Kimi K3 — 2.8T MoE, vision, 1M context
Generate synthetic temporal-awareness training data for LLMs