Decision Index
A benchmark leaderboard established specifically for decision-oriented models covering classification, scoring, routing, and structured judgment tasks. Its evaluation metrics differ from those of general-purpose language model leaderboards, focusing on performance assessment for tasks with finite discrete outputs.
Core Facts
Timeline (last 90 days)
Decision Index is a public scoring benchmark used to compare the performance of different decision models horizontally.
According to officially published data, Nimble-v4-max scores 10.5 points higher than Jev 1.13 on the public Decision Index score.
The community-run Decision Index leaderboard shows Perplexity at 62.75 and Jev at 60.11 (v0.3 scoring), but the scoring formula changes between versions, so the two numbers cannot be directly compared.
On the self-reported Cloudflare internal Decision Index 0.2.1 benchmark, clef-omni leads Jev by a wide margin on intent classification and routing tasks (BANKING77, CLINC150); but on tasks requiring multi-step reasoning (GPQA, BBH, MMLU-Pro), Jev leads by 12 to 31 points.
Decision Index is an evaluation benchmark designed specifically for decision-type models, focusing on structured judgment scenarios such as multi-attribute option comparison, risk probability estimation, and ranking consistency
Aplomb 1 scored 44.86 on Decision Index 0.2.1 using the official tooling, ranking first among 4B models on the published leaderboard.
The author claims that TinyDecide is the smallest model on the Decision Index leaderboard.
Decision Index is a benchmark leaderboard specifically targeting decision-making tasks, and its evaluation metrics differ from those of general language model leaderboards such as MMLU and HellaSwag.
All Facts (5)
Artificial Analysis是一个独立的AI模型基准测试平台，综合编程能力、数学推理、语言理解、知识问答等多个维度评估模型表现
70%UnverifiedDecision Index is a public scoring benchmark used to compare the performance of different decision models horizontally.
50%UnverifiedOn the self-reported Cloudflare internal Decision Index 0.2.1 benchmark, clef-omni leads Jev by a wide margin on intent classification and routing tasks (BANKING77, CLINC150); but on tasks requiring multi-step reasoning (GPQA, BBH, MMLU-Pro), Jev leads by 12 to 31 points.
50%UnverifiedDecision Index is an evaluation benchmark designed specifically for decision-type models, focusing on structured judgment scenarios such as multi-attribute option comparison, risk probability estimation, and ranking consistency
50%UnverifiedDecision Index is a benchmark leaderboard specifically targeting decision-making tasks, and its evaluation metrics differ from those of general language model leaderboards such as MMLU and HellaSwag.
50%