SecIT Bench (Pydantic AI): leaderboard
Cribl's benchmark for IT and security operations agents, run under a plain Pydantic AI harness. Thirty telemetry scenarios across intrusion and breach, covert egress, service errors and performance degradation are each rolled out three times; an incomplete answer scores zero, and the reported figure is the mean scenario accuracy.
Metric: Accuracy (%). Source: secitbench.cribl.io. Status: saturation imminent. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GLM-5.3 | 81.44 |
| 2 | Grok 4.6 | 80.65 |
| 3 | Claude Opus 5 | 79.15 |
| 4 | Qwen 3.8 Max | 78.52 |
| 5 | Muse Spark 1.3 | 77.53 |
| 6 | Kimi K3 | 76.99 |
| 7 | GLM-5.3 Flash | 75.48 |
| 8 | GPT-5.6 Sol | 75.4 |
| 9 | Grok 4.5 | 74.88 |
| 10 | GPT-5.6 Terra | 69.38 |
| 11 | Muse Spark 1.2 | 69.35 |
| 12 | DeepSeek V4 Flash (0731) | 69.02 |
| 13 | Gemini 3.8 Flash | 69 |
| 14 | Claude Opus 4.8 | 67.04 |
| 15 | Gemini 3.7 Flash | 65.12 |
Interactive version: theaggregate.ai/benchmark?slug=secit-bench-pydantic-ai · How It Works · Data refreshed daily, snapshot 2026-09-05.