Kimi K3: second only to Fable 5 on AA-Briefcase

Kimi K3, Moonshot AI's new 2.8 trillion parameter model, ranks second only to Claude Fable 5 on the AA-Briefcase agentic knowledge work benchmark with an Elo score of 1543, demonstrating strong analytical capabilities but weaker presentation quality. However, the model is expensive to operate at $10.57 per task and takes an average of 56.4 minutes per task due to high token usage and multiple reasoning turns, making it less efficient than competing models despite its strong performance metrics.