Artificial Analysis launches Decision Index ranking decision models by capability, cost, speed

Artificial Analysis introduces the Decision Index v1.0, a benchmark that compares decision models on classification, tool calling, knowledge and reasoning, and publishes its methodology and results.

The firm says the index measures capability, cost and speed, and also compares decision models with language models.

In the results, Jev 1.13 from typesafeai scores 58, ahead of GLiDE from Fastino and Microsoft-Decision-1 from Microsoft, which each score 54. Celeris-1 Decision is the fastest at 0.06 seconds per task. Artificial Analysis says Claude Haiku 5.5 (Low), a language model, scores higher at 61, but costs $184 per 1M tasks against $42 for Jev 1.13 and takes 11 times longer per task.