CESBench - Side-Channel: leaderboard
Metric: Composite score (%; mean over the four task types of the item-score means within the D1 side-channel sub-domain (80 items); 380 expert-written items on cryptographic engineering security for IoT devices; zero-shot, temperature 0, no tools, provider default reasoning mode; judgment and scenario answers scored by a Qwen3.5-397B judge from outside the evaluated set). Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.7 Flash | 85.5 |
| 2 | GPT-5.6 Luna | 85 |
| 3 | GLM-5.2 | 84.1 |
| 4 | Kimi K2.6 | 83.5 |
| 5 | DeepSeek V4 Pro | 78.9 |
| 6 | MiniMax-M2.5 | 74.2 |
| 7 | DeepSeek V3 | 63.5 |
| 8 | Ling-flash-2.0 | 61.6 |
| 9 | GLM-4 32B (0414) | 58.2 |
| 10 | Llama 4 Maverick | 57.3 |
| 11 | Hunyuan A13B-Instruct | 47.1 |
Interactive version: theaggregate.ai/benchmark?slug=cesbench-side-channel · How It Works · Data refreshed daily, snapshot 2026-09-26.