List your company

LLM Routing on Agentic Evaluation (test)

89.65AccuracyOracle
59.781267.535675.2983.0444May 8, 2026

Evaluation results

MethodLinks
OracleType=Upper Bound, Model Basket=Earlier2026.0589.6512.7
GPT-4.1Type=Single LLM, Model Basket=Earlier2026.0583.2273.6
DistilBERT (ours)Type=Agent Router, Model Basket=Earlier2026.0582.738.7
GPT-4.1-miniType=Single LLM, Model Basket=Earlier2026.0582.1810.5
GPT-5Type=Single LLM, Model Basket=Earlier2026.0579.1849.9
LengthType=Heuristic Router, Model Basket=Earlier2026.0578.9512.8
GPT-5-nanoType=Single LLM, Model Basket=Earlier2026.0578.81.8
Chat-fine-tunedType=Chat Router, Model Basket=Earlier2026.0577.477.7
GPT-5-miniType=Single LLM, Model Basket=Earlier2026.0577.248
GPT-4.1-nanoType=Single LLM, Model Basket=Earlier2026.0576.673.2
Num. Tool CallsType=Heuristic Router, Model Basket=Earlier2026.0567.8466.9
GPT-5-chatType=Single LLM, Model Basket=Earlier2026.0562.7622.7
Num. TurnsType=Heuristic Router, Model Basket=Earlier2026.0562.6754
o4-miniType=Single LLM, Model Basket=Earlier2026.0560.9323.6