Complexity Ceiling Benchmark - Symbolic Pointer Tracking: leaderboard
The symbolic domain of the Complexity Ceiling Benchmark (Chapra et al., 2026): the model maintains seven variables holding distinct digits through 5 to 50 cyclic-shift and modular-arithmetic updates under an assignment-uniqueness constraint and must reproduce every intermediate state. Accuracy in percent over 400 seeded trials; higher is better.
Metric: Accuracy (%; share of parseable trials whose final answer and every intermediate step are exact, 40 seeded trials at each depth N = 5, 10, ..., 50; temperature 0 via OpenRouter; D2 Symbolic Pointer Tracking, 400 trials per model). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet | 71.2 | #455 |
| 2 | Gemini 2.0 Flash (001) | 30 | #724 |
| 3 | Llama 3.3 70B Instruct | 4.3 | #1353 |
| 4 | GPT-4o Mini | 1.2 | #1627 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=complexity-ceiling-benchmark-symbolic-pointer-tracking · How It Works · Data refreshed daily, snapshot 2026-11-12.