Workflow Execution (Average) on T1–T4 Average benchmark leaderboard · SOTA2 Research

Workflow Execution (Average) on T1–T4 Average

100TSR8 models (Gemma4:e4b/26b/31b, GPT-OSS:20b/120b, MSmall3.2:24b, GPT-5.2)

Evaluation Results

Method
8 models (Gemma4:e4b/26b/31b, GPT-OSS:20b/120b, MSmall3.2:24b, GPT-5.2)2026.05	100	100	100
Qwen2.5:32b2026.05	100	100	100
GPT-4.12026.05	100	100	99.99
Qwen3:32b2026.05	99.98	99.99	99.97
Qwen3:14b2026.05	99.92	99.96	99.91
Qwen3:8b2026.05	99.9	99.95	99.89
Llama3.1:70b2026.05	99.72	99.69	99.7
Qwen2.5:14b2026.05	99.32	99.64	99.56
Llama3.1:8b2026.05	98.02	98.81	98.03
Magistral:24b2026.05	95.6	97.65	95.52
Qwen2.5:7b2026.05	77.54	85.12	77.09