List your company

Realistic Shopping Tasks on WebShop

35.6Success Rate (%)Cloud-only
2.52811.11419.728.286May 23, 2026

Evaluation results

MethodLinks
Cloud-onlyCloud model=Qwen-Max2026.0535.614.7181.5
HeraRouting method=Dynamic routing2026.0534.18.1112.2
FrugalGPTRouting method=Dynamic routing2026.0531.812.1125.8
RouteLLMRouting method=Dynamic routing2026.0531.17.2155.1
Hybrid LLMRouting method=Dynamic routing2026.0529.17.7109.3
Rand. (0.5)Routing probability=0.52026.0527.68.3108.1
EagleRouting method=Dynamic routing2026.0526.68116.6
Rand. (0.4)Routing probability=0.42026.0524.56.997.6
Rand. (0.3)Routing probability=0.32026.0521.25.885.9
Device-onlyOn-device model=Qwen2.5-7B-Instruct2026.053.8061.8