Model comparison
DeepSeek V4 Pro vs Nemotron 3 Ultra
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 42.5 on the Noometry Index.
Last verified . 30 shared benchmarks.
DeepSeek
NVIDIASummary
- They share 30 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 8 categories and Nemotron 3 Ultra in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Pro leads 64.8 to 35.3.
- The biggest single-benchmark swing is ProofBench: 50% for DeepSeek V4 Pro and 2% for Nemotron 3 Ultra.
- Nemotron 3 Ultra is cheaper at $0.50 / $2.20 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 262K.
Side by side
| DeepSeek V4 Pro | Nemotron 3 Ultra | |
|---|---|---|
| Provider | DeepSeek | NVIDIA |
| Noometry Index | 54.3 | 42.5 |
| Released | 2026-04-24 | 2026-06-04 |
| Weights | Open | Open |
| Context window | 1M | 262K |
| Max output | 393K | 128K |
| Input $ / M tokens | $0.66 | $0.50 |
| Output $ / M tokens | $1.98 | $2.20 |
| Results tracked | 48 | 30 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), Nemotron 3 Ultra: 38.1 (#182)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| FrontierCode | 28.6% | 13.6% |
| SciCode | 51% | 40.3% |
| WeirdML | 66.2% | 43.5% |
| LMArena Coding | 1470 | 1468 |
| SWE-bench Verified | 77.6% | — |
| LMArena WebDev | 1582 | — |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use DeepSeek V4 Pro leads
DeepSeek V4 Pro: 32.8 (#58), Nemotron 3 Ultra: 23.4 (#126)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| APEX-Agents | 47.3% | 22.7% |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), Nemotron 3 Ultra: 31.0 (#82)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| CritPt | 18% | 3.1% |
| Chess Puzzles | 47% | 12% |
| LMArena Hard Prompts | 1461 | 1452 |
| Mystery Game Puzzles | 43% | 20% |
| DTBench | 93.9% | 90.1% |
| LMCA | 45.5% | 36.9% |
| Epoch Capabilities Index | 155.31 | 146.17 |
| ARC-AGI-2 | 61.3% | — |
| Kagi LLM Benchmark | 53.5% | — |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 90.5% | — |
| Surface Evolver Bench | 40% | — |
| ForecastBench | 56.1 | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), Nemotron 3 Ultra: 35.3 (#187)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 86.7% |
| ProofBench | 50% | 2% |
| LMArena Math | 1455 | 1457 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), Nemotron 3 Ultra: 52.5 (#61)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| GPQA Diamond | 91.7% | 85.4% |
| LMArena Expert | 1464 | 1472 |
| SimpleQA Verified | 52.9% | — |
| Vectara Hallucination Rate | 8.6% | — |
Multilingual Too close to call
DeepSeek V4 Pro: 54.4 (#45), Nemotron 3 Ultra: 53.5 (#64)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| LMArena Non-English | 1439 | 1427 |
| LMArena Chinese | 1486 | 1497 |
| LMArena French | 1472 | 1472 |
| LMArena German | 1458 | 1471 |
| LMArena Korean | 1447 | 1386 |
| LMArena Russian | 1453 | 1417 |
| LMArena Spanish | 1458 | 1454 |
| LMArena Japanese | 1445 | — |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Pro: 76.1 (#47), Nemotron 3 Ultra: 74.7 (#88)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| LMArena Instruction Following | 1448 | 1418 |
Long Context DeepSeek V4 Pro leads
DeepSeek V4 Pro: 45.0 (#51), Nemotron 3 Ultra: 43.9 (#81)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| LMArena Longer Query | 1458 | 1437 |
| CL-bench Life | 13.5% | — |
Writing & Preference Too close to call
DeepSeek V4 Pro: 65.5 (#46), Nemotron 3 Ultra: 66.3 (#36)
| Benchmark | DeepSeek V4 Pro | Nemotron 3 Ultra |
|---|---|---|
| LMArena Text | 1451 | 1445 |
| LMArena Creative Writing | 1446 | 1398 |
| EQ-Bench Creative Writing | 1553 | 1692 |
| LMArena Multi-Turn | 1467 | 1411 |
| EQ-Bench 4 | 1166 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than Nemotron 3 Ultra?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 42.5 on the Noometry Index.
Which is cheaper, DeepSeek V4 Pro or Nemotron 3 Ultra?
Nemotron 3 Ultra is cheaper. It lists at $0.50 per million input tokens and $2.20 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.
Is DeepSeek V4 Pro or Nemotron 3 Ultra better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 38.1 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Pro does, with 1M tokens against 262K.
How many benchmarks do DeepSeek V4 Pro and Nemotron 3 Ultra share?
30 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Nemotron 3 Ultra has 30.