CISA Dense - RULER (128K): leaderboard
CISA v1 AppendixA Table5a RULER 128K, original Nano30B-A3B and Ultra550B-A55B BF16 dense controls.100 examples per each of13 subtasks, temperature0.00001 and top-p0.99. Sparse budgets and Quest-bound substitutions are separate interventions. Reported metric: Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nemotron 3 Ultra | 97.69 | #245 |
| 2 | NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 | 94.61 | #688 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cisa-dense-ruler-128k · How It Works · Data refreshed daily, snapshot 2026-11-05.