List your company

ResearchFormerly WizWand

LLM Routing

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

0.918nAUC

37

May 13, 2026

Heterogeneous four-agent system Gemma, Phi, Qwen (test)

94Accuracy

27

Mar 4, 2026

CEVAL and GSM8K (OOD)

87.46Performance

21

Feb 26, 2026

MMLU, CMMLU, etc. In-distribution

80.69Performance

21

Feb 26, 2026

0.023Cost (Cost-first)

15

Apr 13, 2026

Agentic Evaluation (test)

89.65Accuracy

14

May 11, 2026

Average across Benchmarks (val)

83Avg Top-1 Acc

14

Feb 26, 2026

66.4Top-1 Acc

14

Feb 26, 2026

96.3Top-1 Acc

14

Feb 26, 2026

SUPERGPQA (val)

0.776Top-1 Acc

14

Feb 26, 2026

91.5Top-1 Acc

14

Feb 26, 2026

66.4Top-1 Accuracy

14

Feb 26, 2026

96.3Top-1 Acc

14

Feb 26, 2026

77.6Top-1 Acc

14

Feb 26, 2026

91.5Top-1 Acc

14

Feb 26, 2026

MMLU-PRO, SUPERGPQA, MEDMCQA, BBEH (test)

91.5MMLU-PRO Top-1 Acc

14

Feb 26, 2026

1.66QNC

14

Apr 20, 2026

In-domain datasets Cost First, alpha=0.8

93Accuracy

11

Mar 4, 2026

In-domain datasets Balance, alpha=0.5

93Accuracy

11

Mar 4, 2026

In-domain datasets Performance First, alpha=0.2

93Accuracy

11

Mar 4, 2026

89Accuracy

11

Mar 4, 2026

OOD datasets (test)

89Accuracy

11

Mar 4, 2026

RouterArena (Evaluation set)

80.72Arena S Score

9

Jun 1, 2026

Mixed-domain ShareGPT, WildChat, Chatbot Arena 10k episodes (test)

914Average Cost

9

May 26, 2026

Six datasets Average

AccMax-Unconstrained

75.9Macro Accuracy

9

May 13, 2026