CoDyControlBench (Success-Stop): leaderboard
CoDyControlBench (2026) under its Success-Stop protocol: the model designs a feedback controller for each of 132 simulated dynamical systems and stops at the first design meeting the steady-state and overshoot criteria, with up to three attempts. Percentage of successful designs; higher is better.
Metric: Design success rate (%; Success-Stop protocol: iteration stops at the first design meeting the criteria, up to three attempts with feedback; 132 controller-design configurations (1-6 DoF, LTI/LTV/nonlinear systems, coupling, damping, SMC- and PID-family controllers), mean of three independent runs; a design succeeds when every response channel meets the steady-state band and overshoot limits). Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (High) | 91.7 | #49 (GPT-5.5) |
| 2 | Gemini 3.1 Pro (Preview) | 91.4 | #94 |
| 3 | GLM-5.2 | 80.3 | #138 |
| 4 | Claude Opus 4.8 | 58.7 | #56 |
| 5 | DeepSeek V4 Pro (High) | 53.3 | #188 (DeepSeek V4 Pro) |
| 6 | Qwen 3.5 397B A17B | 46.8 | #246 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=codycontrolbench-success-stop · How It Works · Data refreshed daily, snapshot 2026-11-13.