Nvidia Nemotron 3 Ultra 550B
Nvidia's Nemotron 3 Ultra 550B A55B model from the Nemotron 3 family. It uses a hybrid Mamba-Transformer MoE architecture. Provider-specific context limits vary, with the longest current route supporting up to 1M context.
- Reasoning
- Tool Calling
- Parallel Tool Calls
Added Jun 4, 2026
Model weightsPricing
Auto routing · per 1M tokens- Input
- $0.50
- Output
- $2.50
- Cache read
- $0.15
Specifications
- Context window
- 512.3K
- Max output
- 65.5K
- Parameters
- 550B / 55B
- Total / active
Benchmarks
Benchmarks
Sourced from LMArena.
Arena Score
1426.3
Overall Rank
#123 / 421
Votes
11,392
Confidence Interval
1419.4 - 1433.2
Category Scores
Coding
#124 / 416
3,146 votes
1473.7
Math
#92 / 404
574 votes
1442.3
Longer Query
#128 / 399
5,228 votes
1432.6
Creative Writing
#145 / 419
2,148 votes
1382.4
Instruction Following
#143 / 421
3,839 votes
1405.5
Hard Prompts
#130 / 421
7,482 votes
1443.8
Additional Categories21
German
#85 / 320
211 votes
1447.1
Industry Mathematical
#92 / 402
703 votes
1451.3
Expert
#101 / 371
1,218 votes
1466.9
French
#112 / 301
458 votes
1447.9
Spanish
#113 / 304
452 votes
1427.4
Chinese
#121 / 395
718 votes
1463.9
English
#122 / 421
5,294 votes
1439.0
Exclude Ties
#123 / 421
8,659 votes
1417.0
Industry Life And Physical And Social Science
#123 / 419
1,856 votes
1446.4
Industry Software And It Services
#124 / 421
4,364 votes
1463.6
Non English
#131 / 421
6,097 votes
1406.0
Industry Business And Management And Financial Operations
#135 / 414
2,254 votes
1418.7
Industry Legal And Government
#135 / 393
950 votes
1428.2
Hard Prompts English
#137 / 419
3,616 votes
1446.7
Industry Medicine And Healthcare
#140 / 391
844 votes
1435.2
Industry Writing And Literature And Language
#143 / 420
2,996 votes
1392.9
Russian
#144 / 385
1,174 votes
1404.8
Industry Entertainment And Sports And Media
#145 / 419
2,589 votes
1380.4
Korean
#146 / 289
203 votes
1348.7
Multi Turn
#174 / 419
1,760 votes
1396.0
Polish
#188 / 243
224 votes
1345.7
Published 2026-11-11 · Matched as NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) · CC BY 4.0
LMArena DatasetLeaderboard results
| Benchmark | Score | Rank on NanoGPT |
|---|---|---|
Epoch AI Benchmarking Hub Epoch AI data, CC BY 4.0 · Updated Nov 13, 2026 | ||
GPQA Diamond | 85.4% | #67 of 119 |
OTIS Mock AIME 2024–2025 | 86.7% | #66 of 121 |
Mystery Game Puzzles | 20.0% | #48 of 77 |
Filtered and ranked to models on NanoGPT.
Providers
Choose or restrict providers with the API. Provider-selection prices include the selection markup. Provider selection guide · Pricing API (JSON, USD per 1K tokens)
 | $0.50 | $2.50 | $0.15 | N/A | N/A |
| $0.50 | $2.50 | $0.15 | N/A | N/A |
| $0.105 | $3.15 | $0.105 | N/A | N/A | |
| $0.525 | $2.31 | $0.105 | N/A | N/A | |
| $0.63 | $3.78 | $0.21 | N/A | N/A | |
| $0.689 | $3.45 | N/A | N/A | N/A |
Batch rows for this model use normal routing and cost the same as standard API requests. Set provider in a row to choose a provider.