EnterpriseOps-Gym-AA - CSM: leaderboard
Artificial Analysis's implementation of ServiceNow's EnterpriseOps-Gym restricted to the customer service management (CSM) domain: stateful, multi-step enterprise workflows completed through live tool use and graded on the final database state, with no partial credit. Task success rate, 0-100 (%); higher is better.
Metric: Task Success (%). Source: artificialanalysis.ai. Saturation forecast: Around December 2027. 50 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 55.38 | #22 (GPT-5.6 Sol) |
| 2 | Grok 4.6 (High) | 55.38 | #29 (Grok 4.6) |
| 3 | GLM-5.3 (Max) | 54.84 | #44 (GLM-5.3) |
| 4 | GPT-5.5 (xHigh) | 53.94 | #33 (GPT-5.5) |
| 5 | Claude Opus 5 (Max) | 53.41 | #16 (Claude Opus 5) |
| 6 | Gemini 3.5 Flash (High) | 52.69 | #66 (Gemini 3.5 Flash) |
| 7 | DeepSeek V4 Pro (0813) (Max) | 51.61 | #58 (DeepSeek V4 Pro (0813)) |
| 8 | Gemini 3.7 Flash (Medium) | 50.54 | #24 (Gemini 3.7 Flash) |
| 9 | GPT-5.5 (Medium) | 50.36 | #33 (GPT-5.5) |
| 10 | GPT-5.6 Luna (Max) | 49.46 | #79 (GPT-5.6 Luna) |
| 11 | Grok 4.5 (High) | 49.46 | #54 (Grok 4.5) |
| 12 | GPT-5.6 Terra (Max) | 48.92 | #40 (GPT-5.6 Terra) |
| 13 | Step 5 Preview | 48.92 | #53 |
| 14 | Claude Sonnet 5 (Max) | 48.92 | #65 (Claude Sonnet 5) |
| 15 | Claude Sonnet 5 (xHigh) | 48.03 | #65 (Claude Sonnet 5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=enterpriseops-gym-aa-csm · How It Works · Data refreshed daily, snapshot 2026-11-05.