CESBench - Integration: leaderboard

Metric: Composite score (%; mean over the four task types of the item-score means within the D6 integration and key management sub-domain (53 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

#ModelScore
1GLM-5.290.7
2Kimi K2.689.7
3DeepSeek V4 Pro87.6
4MiniMax-M2.586.5
5GPT-5.6 Luna84.3
6Gemini 3.7 Flash81.5
7DeepSeek V379.3
8Llama 4 Maverick73.8
9Ling-flash-2.073.5
10GLM-4 32B (0414)68.9
11Hunyuan A13B-Instruct64.9

Interactive version: theaggregate.ai/benchmark?slug=cesbench-integration · How It Works · Data refreshed daily, snapshot 2026-09-26.