SecRespond - Baseline Risk Detection: leaderboard
Metric: Detection CAP-score (%) on the Baseline Risk (BAS) capability dimension: the detection (discovery, evidence and root-cause attribution, 0-3 per checkpoint) score achieved over all checkpoints mapped to the dimension across the 10 SecRespond cyber ranges (averaged over three LLM judges); agents in the OpenCode harness with reasoning enabled; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 (Thinking) | 84.2 |
| 2 | GLM-5.1 | 81.1 |
| 3 | Qwen 3.7 Max (Thinking) | 80.9 |
| 4 | Claude Opus 4.6 (Thinking) | 78.6 |
| 5 | Claude Opus 4.5 (Thinking) | 75.8 |
| 6 | Qwen 3.6 Plus (Thinking) | 74.7 |
| 7 | GPT-5.4 Pro (xHigh) | 73.5 |
| 8 | MiniMax-M2.7 | 68.6 |
| 9 | GLM-5 | 67.8 |
| 10 | Kimi K2.6 (Thinking) | 66.7 |
| 11 | Claude Sonnet 4.5 (Thinking) | 66.2 |
| 12 | GPT-5.5 | 63.6 |
| 13 | GPT-5.2 Pro | 62.4 |
| 14 | Qwen 3.5 Plus (Thinking) | 61.9 |
| 15 | GPT-5.4 | 61.8 |
Interactive version: theaggregate.ai/benchmark?slug=secrespond-baseline-risk-detection · How It Works · Data refreshed daily, snapshot 2026-09-29.