Artificial Analysis Decision Index v2 (artificialanalysis.ai)

On November 8, 2026, the Artificial Analysis Decision Index v2 was launched, introducing a revamped evaluation framework for decision models in AI. This update is crucial for the AI/ML community as it emphasizes real-world application and improves protection against strategies that exploit benchmark weaknesses. Notably, 45% of the new index is derived from private datasets, enhancing its robustness. The Decision Score metric is a key addition, rewarding models not just for correct answers but also for calibrated confidence in their predictions, creating a comprehensive evaluation system that places more emphasis on practical performance. The index reveals significant competition among leading decision models. Drex Max v1.1 from Nace AI takes the lead with a score of 73, trailed closely by GLiDE v1.2 and Mammoth 27B, both scoring 72. This update also highlights varied costs associated with executing tasks—ranging from as low as $14 to as high as $859 per million tasks—illustrating a wide disparity in operational efficiency across models. Furthermore, advanced models like Claude Haiku 5.5 and GPT-6 Luna achieve higher decision scores, albeit at significantly increased costs and processing times, indicating a trade-off between performance and resource expenditure. This multidimensional evaluation could inform future model development and deployment strategies in the fast-evolving landscape of AI technologies.