Nemotron 3 Ultra
NVIDIA's largest Nemotron 3 model.
NVIDIA, reasoning, open weights. Checked 30 October 2026.
Where to use it through an API
NVIDIAnvidia/nemotron-3-ultra-550b-a55bFree key
The maker publishes its own benchmarks and limits on its model page.
How it compares
The models just above and below it on the Artificial Analysis Intelligence Index.
| Model | Index | Context | Speed |
|---|---|---|---|
| Mistral Large 4 | 26 | 1M | 161/s |
| MiniMax M3 | 19 | 1M | 110/s |
| Gemini 3.5 Flash-Lite | 12 | 1M | 454/s |
| Nemotron 3 Ultra | 12 | 262K | 158/s |
Use Nemotron 3 Ultra with every other AI at once
Put Nemotron 3 Ultra in one panel and ChatGPT, Claude, Gemini or any of 300+ models in the others. One message goes to all of them, answers arrive side by side, and test mode times them and hides the names so you judge fairly. Free during the beta with your own keys.
Questions
Is there a free Nemotron 3 Ultra API?
Yes, within limits: NVIDIA give free keys that reach Nemotron 3 Ultra.
What is the Nemotron 3 Ultra model id?
NVIDIA: nvidia/nemotron-3-ultra-550b-a55b.
How does Nemotron 3 Ultra score?
12 on the Artificial Analysis Intelligence Index, checked 30 October 2026.
Can I use Nemotron 3 Ultra with ChatGPT, Claude and Gemini at the same time?
Yes. MultiScreenAI puts Nemotron 3 Ultra in one panel and other models in the others, sends one message to all of them and shows the answers side by side.