Nemotron 3 Ultra
Intelligence benchmarks
Artificial Analysis indexes - compared with the best open and proprietary models
Intelligence
47.7
AA Index
Coding
37.6
AA Index
Agentic
57.1
AA Index
Intelligence Index - Nemotron 3 Ultra vs. the field
Best open-weight models (you can run locally) and leading proprietary models for context.
Coding Index comparison
Agentic Index comparison
Benchmark data from Artificial Analysis · updated 2026-06-07.
Standard benchmarks
Performance across standard evaluations
| Benchmark | Score |
|---|---|
| GPQA | 86.7 |
Will it run on your hardware?
Pick your GPU memory - see which quantizations fit, and the cheapest card for the rest
Need an exact figure for your context length? Use the VRAM calculator.
Run it locally
Copy-paste - running in under a minute
vllm serve nvidia/Llama-3_1-Nemotron-Ultra-253B-v1-FP8New to this? Start with Ollama · serve to many users with vLLM.
Deep dive
Notes, sources, and the full write-up
Nemotron 3 Ultra
Nemotron 3 Ultra is a 253.4B-parameter other model from NVIDIA. It scores 47.7 on the Artificial Analysis Intelligence Index (coding 37.6). At Q4_K_M it needs roughly 147 GB of VRAM, placing it in the 48 GB+ / multi-GPU hardware tier.
Specifications
| Spec | Value |
|---|---|
| Parameters | 253.4B |
| Context length | 131K tokens |
| License | other |
| Modalities | text |
| Released | 2025-04-30 |
| Weights | nvidia/Llama-3_1-Nemotron-Ultra-253B-v1-FP8 |
Benchmarks
Artificial Analysis Intelligence Index - Nemotron 3 Ultra vs. leading closed models:
| Model | Intelligence | Coding | GPQA |
|---|---|---|---|
| Nemotron 3 Ultra | 47.7 | 37.6 | 86.7 |
| GPT-5.5 (xhigh) | 60.2 | 59.1 | 93.5 |
| Claude Opus 4.8 (max) | 61.4 | 56.7 | 92 |
| Gemini 3.1 Pro Preview | 57.2 | 55.5 | 94.1 |
| Grok 4.3 (high) | 53.2 | 41 | 90.1 |
Source: Artificial Analysis (2026-06-04).
VRAM requirements
| Quant | VRAM | Runs on |
|---|---|---|
| Q4_K_M | ~147 GB | multi-GPU / datacenter |
| Q5_K_M | ~180 GB | multi-GPU / datacenter |
| Q8_0 | ~271 GB | multi-GPU / datacenter |
| FP16 | ~507 GB | multi-GPU / datacenter |
VRAM is estimated from parameter count; MoE models still need all weights resident.
How to run
vLLM:
vllm serve nvidia/Llama-3_1-Nemotron-Ultra-253B-v1-FP8Popularity
Nemotron 3 Ultra has 9,274 downloads in the last month on HuggingFace and 12 likes.
Frequently asked
Quick answers to common questions
How much VRAM does Nemotron 3 Ultra need?
Nemotron 3 Ultra with 253.4B parameters needs approximately 147 GB at Q4_K_M quantization. Use our VRAM calculator for an exact estimate.
Is Nemotron 3 Ultra better than other nvidia models?
Nemotron 3 Ultra has 253.4B parameters with 131,072 context - a strong choice for general use.
What license is Nemotron 3 Ultra under?
Nemotron 3 Ultra is released under the other license, making it suitable for most commercial and personal projects.
What hardware runs Nemotron 3 Ultra well?
With 253.4B parameters, Nemotron 3 Ultra requires adequate VRAM. High-end GPUs like the RTX 4090 (24GB), RTX 5090 (32GB), or Mac Studio with unified memory are good options. Check our hardware directory for specific recommendations.
What is the best quantization for Nemotron 3 Ultra?
Q4_K_M is the recommended sweet spot - ~98% of FP16 quality at ~27% of the size. Q5_K_M (~180 GB) is an option if you have spare VRAM. Use our VRAM calculator to compare.
How long can Nemotron 3 Ultra's context window handle?
Nemotron 3 Ultra supports a 131,072-token context window - enough for very long documents, codebases, or multi-turn conversations. Real-world usable context may vary by implementation.
What models compete with Nemotron 3 Ultra?
Nemotron 3 Ultra competes with other 127B–380B. Browse our model directory for comparisons, benchmarks, and community reviews to find the best fit.
Nearby options
Similar models and compatible hardware by spec
Similar by size
Comments coming soon
Configure NEXT_PUBLIC_GISCUS_REPO_ID and NEXT_PUBLIC_GISCUS_CATEGORY_ID at giscus.app to enable.