Benchmarks / Customer service

Customer service

Serve customers with tools while following policy.

Part of Customer support.

• AutomationBenchby ZapierTask level · Pass rate (task_completed_correctly)

  51.3%Gemini 4 Argon (High) · 30 Sep 2026

  Emerging
• CRMArena-Proby Salesforce AI ResearchTask level · Single-turn task success (B2C org)

  58.3%gemini-2.5-pro (ReAct) · 24 May 2025

  Emerging
• EnterpriseOps-Gymby ServiceNow, Mila and Université de MontréalTask level · Task success rate

  45.9%Claude Opus 4.6 · 16 Mar 2026

  Emerging
• τ²-benchby SierraTask level · Telecom pass^1

  97.8%Qwen3.5-397B-A17B (thinking) · 27 Feb 2026

  Saturated