New decision-model index finds speed and cost gains, but weaker reasoning

Decision models are quicker and cheaper than language models in Artificial Analysis’s new benchmark, but they currently lag on knowledge and reasoning. Its first Decision Index puts Jev 1.13 ahead of GLiDE and Microsoft-Decision-1, offering a useful early map of a fast-growing corner of AI.

Watch Desk analysis

What happened

Artificial Analysis launched version 1.0 of its Decision Index on 13 October. It evaluates decision models across classification, tool calling, knowledge, hallucination and reasoning tasks. The index ranks Jev 1.13 first, ahead of GLiDE and Microsoft-Decision-1.

The index says decision models have speed and cost advantages over language models, while trailing them in knowledge and reasoning. That is the benchmark publisher’s account of its results, not a universal verdict on every model or task.

Why it matters

Decision models are built for structured jobs such as classification and tool use, rather than open-ended conversation. A comparison across both strengths and weaknesses can help teams decide where this narrower approach fits, and where it may need a language model alongside it.

The trade-off matters: a cheaper, faster system is not an obvious bargain if the task depends on knowledge or reasoning. The index gives buyers a starting point for that decision, not a substitute for testing models on their own workloads.

Our read

This is a useful first scoreboard for a category that risks being judged on speed alone. The ranking is worth watching, but the more revealing result may be the split between efficiency and capability. A model can win the stopwatch and still lose the harder question.

What to watch

• Whether later editions add more models and broaden the tested tasks.
• How the benchmark’s results compare with teams’ own workload tests.
• Whether decision models narrow their knowledge and reasoning gaps.

Discussion spark: For a structured task, would you choose a faster, cheaper decision model with weaker reasoning, or pay more for a broader language model?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.