Model comparison
Nemotron 3 Ultra vs Step 5 Preview
Step 5 Preview is the stronger model overall, scoring 47.9 to 42.5 on the Noometry Index. Nemotron 3 Ultra costs 1.5× less per token, which makes it the better buy when Step 5 Preview's lead doesn't matter for your workload.
Last verified . 16 shared benchmarks.
NVIDIA
StepFunSummary
- They share 16 benchmarks with published results for both. Nemotron 3 Ultra scores higher in 2 categories and Step 5 Preview in 6 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in coding, where Step 5 Preview leads 51.8 to 38.1.
- The biggest single-benchmark swing is ProofBench: 2% for Nemotron 3 Ultra and 42% for Step 5 Preview.
- Nemotron 3 Ultra is cheaper at $0.50 / $2.20 per million input/output tokens, against $1 / $2.70 for Step 5 Preview.
- Step 5 Preview accepts more context: 1.02M tokens versus 262K.
- Nemotron 3 Ultra has downloadable open weights; the other is API-only.
Side by side
| Nemotron 3 Ultra | Step 5 Preview | |
|---|---|---|
| Provider | NVIDIA | StepFun |
| Noometry Index | 42.5 | 47.9 |
| Released | 2026-06-04 | 2026-09-16 |
| Weights | Open | Proprietary |
| Context window | 262K | 1.02M |
| Max output | 128K | 66K |
| Input $ / M tokens | $0.50 | $1 |
| Output $ / M tokens | $2.20 | $2.70 |
| Results tracked | 30 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Step 5 Preview leads
Nemotron 3 Ultra: 38.1 (#182), Step 5 Preview: 51.8 (#36)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| SciCode | 40.3% | 58.9% |
| LMArena Coding | 1468 | 1480 |
| FrontierCode | 13.6% | — |
| LMArena WebDev | — | 1564 |
| WeirdML | 43.5% | — |
Agentic & Tool Use Not comparable
Nemotron 3 Ultra: 23.4 (#126), Step 5 Preview: —
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| APEX-Agents | 22.7% | — |
Reasoning Step 5 Preview leads
Nemotron 3 Ultra: 31.0 (#82), Step 5 Preview: 40.0 (#57)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| CritPt | 3.1% | 20.9% |
| LMArena Hard Prompts | 1452 | 1465 |
| Chess Puzzles | 12% | — |
| Mystery Game Puzzles | 20% | — |
| DTBench | 90.1% | — |
| LMCA | 36.9% | — |
| Epoch Capabilities Index | 146.17 | — |
Math Step 5 Preview leads
Nemotron 3 Ultra: 35.3 (#187), Step 5 Preview: 46.1 (#72)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| ProofBench | 2% | 42% |
| LMArena Math | 1457 | 1470 |
| OTIS Mock AIME 2024-2025 | 86.7% | — |
Knowledge Nemotron 3 Ultra leads
Nemotron 3 Ultra: 52.5 (#61), Step 5 Preview: 41.2 (#112)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| LMArena Expert | 1472 | 1470 |
| GPQA Diamond | 85.4% | — |
Multimodal Not comparable
Nemotron 3 Ultra: —, Step 5 Preview: 41.0 (#41)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| LMArena Vision | — | 1267 |
Multilingual Too close to call
Nemotron 3 Ultra: 53.5 (#64), Step 5 Preview: 53.8 (#54)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| LMArena Non-English | 1427 | 1432 |
| LMArena Chinese | 1497 | 1519 |
| LMArena Russian | 1417 | 1436 |
| LMArena Spanish | 1454 | 1439 |
| LMArena French | 1472 | — |
| LMArena German | 1471 | — |
| LMArena Korean | 1386 | — |
Instruction Following Step 5 Preview leads
Nemotron 3 Ultra: 74.7 (#88), Step 5 Preview: 76.0 (#50)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| LMArena Instruction Following | 1418 | 1444 |
Long Context Too close to call
Nemotron 3 Ultra: 43.9 (#81), Step 5 Preview: 44.5 (#64)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| LMArena Longer Query | 1437 | 1455 |
Writing & Preference Nemotron 3 Ultra leads
Nemotron 3 Ultra: 66.3 (#36), Step 5 Preview: 62.8 (#71)
| Benchmark | Nemotron 3 Ultra | Step 5 Preview |
|---|---|---|
| LMArena Text | 1445 | 1442 |
| LMArena Creative Writing | 1398 | 1410 |
| LMArena Multi-Turn | 1411 | 1447 |
| EQ-Bench Creative Writing | 1692 | — |
Frequently asked questions
Is Nemotron 3 Ultra better than Step 5 Preview?
Step 5 Preview is the stronger model overall, scoring 47.9 to 42.5 on the Noometry Index. Nemotron 3 Ultra costs 1.5× less per token, which makes it the better buy when Step 5 Preview's lead doesn't matter for your workload.
Which is cheaper, Nemotron 3 Ultra or Step 5 Preview?
Nemotron 3 Ultra is cheaper. It lists at $0.50 per million input tokens and $2.20 per million output tokens; Step 5 Preview lists at $1 and $2.70.
Is Nemotron 3 Ultra or Step 5 Preview better for coding?
Step 5 Preview scores higher on coding benchmarks: 51.8 versus 38.1 in the Noometry coding category.
Which has the bigger context window?
Step 5 Preview does, with 1.02M tokens against 262K.
How many benchmarks do Nemotron 3 Ultra and Step 5 Preview share?
16 benchmarks have published results for both models. Nemotron 3 Ultra has 30 scored results on Noometry and Step 5 Preview has 18.