SaaS Marathon Bench · Task 8
Expense recent business trips
Task pass rate35% ±12%
After recent business trips, the agent builds an expense report per trip in a Google Sheet, matching each charge to its receipt, and flags what has no receipt or looks off. It says which meetings travel forced to be moved or missed and logs each as a note on that company's record in Attio. If a supplier needs writing to, it drafts the note.
Results
ModelPass rate
Claude Opus 5.5CLI100%
Gemini 3.8 FlashCLI100%
GPT-6 AstraCLI100%
Claude Opus 5.5GUI67%
Claude Sonnet 5.5CLI67%
GPT-6.1 SolCLI67%
Grok 4.7CLI67%
Claude Haiku 5.5CLI33%
Claude Sonnet 5.5GUI33%
GPT-6.1 SolGUI33%
Qwen3.8 MaxCLI33%
Claude Haiku 5.5GUI0%
DeepSeek V4 FlashCLI0%
Gemini 3.8 FlashGUI0%
Gemini 4 ArgonGUI0%
Gemini 4 ArgonCLI0%
GLM 5CLI0%
GPT-6 AstraGUI0%
Grok 4.7GUI0%
Qwen3.8 MaxGUI0%