NVIDIA Nemotron 49B v1.5 Reproduction - ruler - ruler.nemotron_128k: leaderboard
RULER long-context test at 128k, the 128k aggregate score, the mean of the 13 RULER task scores reported with it; accuracy in percent for reasoning on and off. From NVIDIA's NeMo-Skills tutorial reproducing the Llama-3.3-Nemotron-Super-49B-v1.5 evaluations, using the nemotron_super_128k RULER setup prepared with a 131,072-token maximum sequence length. Reasoning on uses temperature 0.6 and top-p 0.95; reasoning off uses temperature 0, top-p 1.0 and the /no_think system prompt; generation length comes from RULER's predefined setup.
Metric: Accuracy (%). Source: raw.githubusercontent.com. Saturation forecast: Around February 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama-3.3-nemotron-super-49B-v1.5 (Reasoning) | 66.7 | #903 (Llama-3.3-nemotron-super-49B-v1.5) |
| 2 | Llama-3.3-nemotron-super-49B-v1.5 (Non-reasoning) | 66.1 | #903 (Llama-3.3-nemotron-super-49B-v1.5) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=nvidia-nemotron-49b-v1-5-reproduction-ruler-ruler-nemotron-128k · How It Works · Data refreshed daily, snapshot 2026-11-12.