CESBench - Scenario: leaderboard
Metric: Scenario score (%; the 63 open-ended engineering diagnosis items, mean of 3 to 5 expert rubric dimensions judged 0, 0.5 or 1; 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 | GPT-5.6 Luna | 88.4 |
| 2 | GLM-5.2 | 84.2 |
| 3 | Kimi K2.6 | 82.2 |
| 4 | Gemini 3.7 Flash | 81.9 |
| 5 | DeepSeek V4 Pro | 81.4 |
| 6 | MiniMax-M2.5 | 71.9 |
| 7 | DeepSeek V3 | 67.5 |
| 8 | GLM-4 32B (0414) | 59.1 |
| 9 | Llama 4 Maverick | 57.1 |
| 10 | Ling-flash-2.0 | 55.9 |
| 11 | Hunyuan A13B-Instruct | 47.2 |
Interactive version: theaggregate.ai/benchmark?slug=cesbench-scenario · How It Works · Data refreshed daily, snapshot 2026-09-26.