List your company

Task Efficient Routing on Software scenario

0.0138CostForce Weak
-0.0984640.6593181.41712.174882Jan 27, 2026

Evaluation results

MethodLinks
Force WeakModel=qwen2026.010.013865.2—
Force WeakModel=qwen2026.010.01380.65297.8
CASTERModel=qwen2026.010.018653.1—
CASTERModel=qwen2026.010.01860.531100
Force StrongModel=qwen2026.010.0397——
Force StrongModel=qwen2026.010.0397—99.3
Force WeakModel=openai2026.010.049896.6—
Force WeakModel=openai2026.010.04980.96682.2
CASTERModel=deepseek2026.010.110652.9—
CASTERModel=deepseek2026.010.11060.52998.1
Force WeakModel=deepseek2026.010.139140.7—
Force WeakModel=deepseek2026.010.13910.40798.3
Force StrongModel=deepseek2026.010.2346——
Force StrongModel=deepseek2026.010.2346—98.6
CASTERModel=openai2026.010.405272.4—
CASTERModel=openai2026.010.40520.72497
Force WeakModel=gemini2026.010.441270.8—
Force WeakModel=gemini2026.010.44120.70899.8
CASTERStrategy=Dynamic Routing2026.010.5848—
CASTERModel=gemini2026.010.878341.8—
CASTERModel=gemini2026.010.87830.418100
Force WeakModel=claude2026.010.885868.6—
Force WeakModel=claude2026.010.88580.68693.3
FrugalGPT (Cascade)Strategy=Cascade2026.011.11——
CASTERModel=claude2026.011.373851.3—
CASTERModel=claude2026.011.37380.51396.4
Force StrongModel=openai2026.011.4658——
Force StrongModel=openai2026.011.4658—95.3
Force StrongModel=gemini2026.011.5087——
Force StrongModel=gemini2026.011.5087—100
Force StrongModel=claude2026.012.8204——
Force StrongModel=claude2026.012.8204—100