“The AlphaSense study found that for financial analysis tasks, QIMI K3 and GLM 5.2 performed at a similar level to Sonnet 5 and were more expensive than GPT-5.6-SOL.”