Mathematical Reasoning on GSM-Hard (Acc, Lat, SLO)
64.6AccuracyGPTSwarm
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPTSwarmPool=Pure-API, Load Rate (µ)=50 r/m, Budget=500 s2026.06 | 64.6 | 686 | 23.6 | |
| INFRAMINDPool=Pure-API, Load Rate (µ)=100 r/m, Budget=500 s2026.06 | 64.6 | 492 | 50.4 | |
| GPTSwarmPool=Pure-API, Load Rate (µ)=100 r/m, Budget=500 s2026.06 | 64 | 902 | 11.6 | |
| INFRAMINDPool=Pure-API, Load Rate (µ)=50 r/m, Budget=500 s2026.06 | 63.7 | 370 | 68.1 | |
| GPTSwarmPool=Pure-API, Load Rate (µ)=10 r/m, Budget=500 s2026.06 | 63.6 | 70 | 99.8 | |
| MASRouterPool=Pure-API, Load Rate (µ)=50 r/m, Budget=500 s2026.06 | 63.6 | 714 | 40.6 | |
| INFRAMINDPool=Pure-API, Load Rate (µ)=10 r/m, Budget=500 s2026.06 | 63 | 3 | 100 | |
| MASRouterPool=Hybrid, Load Rate (µ)=10 r/m, Budget=100 s2026.06 | 62.8 | 28 | 99.8 | |
| MASRouterPool=Pure-API, Load Rate (µ)=10 r/m, Budget=500 s2026.06 | 62.8 | 114 | 96.2 | |
| INFRAMINDPool=Hybrid, Load Rate (µ)=10 r/m, Budget=100 s2026.06 | 62.4 | 14 | 100 | |
| INFRAMINDPool=Hybrid, Load Rate (µ)=50 r/m, Budget=100 s2026.06 | 61.6 | 43 | 94.4 | |
| MASRouterPool=Hybrid, Load Rate (µ)=50 r/m, Budget=100 s2026.06 | 61.4 | 352 | 17.4 | |
| INFRAMINDPool=Hybrid, Load Rate (µ)=100 r/m, Budget=100 s2026.06 | 61.4 | 107 | 54.4 | |
| MASRouterPool=Pure-API, Load Rate (µ)=100 r/m, Budget=500 s2026.06 | 61.2 | 813 | 30.6 | |
| MASRouterPool=Hybrid, Load Rate (µ)=100 r/m, Budget=100 s2026.06 | 60.4 | 441 | 11.6 | |
| GPTSwarmPool=Hybrid, Load Rate (µ)=50 r/m, Budget=100 s2026.06 | 59.8 | 209 | 19.6 | |
| GPTSwarmPool=Hybrid, Load Rate (µ)=10 r/m, Budget=100 s2026.06 | 59.4 | 42 | 100 | |
| GPTSwarmPool=Hybrid, Load Rate (µ)=100 r/m, Budget=100 s2026.06 | 58.8 | 412 | 9.4 |