Search Intelligence Score
53.0
+37.7 lift from search
050100
- Lift from search
- +37.7
- 15.3 without search
- Cost per 1K tasks
- $113
- Below the median score of 61.9, so not ranked on cost
- Time per task
- 328s
- Search plus inference time to complete one task
Accuracy by benchmark
Accuracy on each eval, with search (solid) and without (light).
- DSQA71.0 vs 27.9
- HLE34.0 vs 10.0
- WISER54.0 vs 8.0
Cost per 1K tasks
How the cost splits between model inference, search calls, and page extraction.
- Inference
- $100
- 89%
- Search
- $6.67
- 6%
- Extract
- $5.75
- 5%
Compared with
Other models from the same provider alongside models that score closest.
- Nem 3 Ultra53.0Cost $113 · Lift +37.7 · 328s
- Nem 3.5 Ltng37.9Cost $12.1 · Lift +28.0 · 176s
- DS Flash53.1Cost $24.4 · Lift +33.7 · 358s
- MiniMax M352.6Cost $41.1 · Lift +29.5 · 291s
- DS Pro 081353.5Cost $234 · Lift +20.4 · 394s
| Model | Score | Lift | Cost per 1K tasks | Time per task |
|---|---|---|---|---|
| Nemotron 3 Ultra 550B | 53.0 | +37.7 | $113 | 328s |
| Nem 3.5 LtngNemotron 3.5 Lightning | 37.9 | +28.0 | $12.1 | 176s |
| DS FlashDeepSeek V4 Flash | 53.1 | +33.7 | $24.4 | 358s |
| MiniMax M3MiniMax M3 | 52.6 | +29.5 | $41.1 | 291s |
| DS Pro 0813DeepSeek V4 Pro (0813) | 53.5 | +20.4 | $234 | 394s |
Search Intelligence Score averages accuracy on DSQA, HLE, and WISER with equal weight; cost and time are averaged the same way, with cost shown per 1,000 tasks. Latest update September 25, 2026. Full methodology.
# Nemotron 3 Ultra 550B · Search Capability Leaderboard
- Search Intelligence Score: 53.0 (#23 in Search Intelligence)
- Without search: 15.3 · Lift from search +37.7
- Cost per 1K tasks: $113 (below the median score of 61.9, so not ranked on cost)
- Time per task: 328s
- Lab: NVIDIA · Model id: nvidia/nemotron-3-ultra-550b-a55b
## Accuracy by benchmark
| Suite | With search | Without search |
|---|---|---|
| DSQA | 71.0 | 27.9 |
| HLE | 34.0 | 10.0 |
| WISER | 54.0 | 8.0 |
## Compared with
| Model | Score | Lift | Cost per 1K tasks | Time per task |
|---|---|---|---|---|
| Nemotron 3 Ultra 550B | 53.0 | +37.7 | $113 | 328s |
| Nemotron 3.5 Lightning | 37.9 | +28.0 | $12.1 | 176s |
| DeepSeek V4 Flash | 53.1 | +33.7 | $24.4 | 358s |
| MiniMax M3 | 52.6 | +29.5 | $41.1 | 291s |
| DeepSeek V4 Pro (0813) | 53.5 | +20.4 | $234 | 394s |
Latest update September 25, 2026. Methodology: /leaderboard#methodology