Nemotron 3 Ultra is NVIDIA's flagship open-weight model in the Nemotron 3 family, a 550B-parameter mixture-of-experts model with 55B active parameters. It is optimized for orchestrating complex, long-running agent workflows, combining frontier reasoning with high inference throughput.
Cost rate
0.5x
Context
262K
Released
Jun 4, 2026
Input
Text
Output
Text
Support
ReasoningTool callingStructured outputs
Performance
Median latency and throughput measured across recent requests.
Best at
The categories where this model ranks highest.
No top-ranked categories found.
| Domain | Score | Rank |
|---|---|---|
| AgenticAgent Arena score | -0.17 | #64 |
| Category | Win rate | Rank |
|---|---|---|
| ASCII Art | 36.3% | #68 |
| SVG | 31.9% | #81 |
| Data Visualization | 37.2% | #90 |
| 3D | 37.9% | #91 |
| Game Development | 35.3% | #91 |
| UI Component | 35.1% | #98 |
| Code Categories | 35.3% | #101 |
| Website | 34.3% | #106 |
Latest news about Nemotron 3 Ultra
Recent articles and community posts, linked to their original sources.
- NVIDIA launches Nemotron for scientific discovery agentsEngineering.com
Nemotron 3.5 Super VL: BEST Free Model for Hermes Agent? (2026 Verdict)Hostinger Academy2,572
Beyond Transcription: Nemotron 3.5 Transcribe | Nemotron LabsNVIDIA Developer3,090
NVIDIA's Nemotron Bolt 74B-A9B 2 BitSam Witteveen69,185
Nvidia Just Released NVIDIA Nemotron 3.5 Transcribe! 🤯Julian Goldie SEO1,604
Nvidia Nemotron 3.5 Transcribe: open ASR for 40 languagesHacker News13426