Model comparison
DeepSeek V4 Flash vs Nemotron 3 Ultra
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 44.7 on the Noometry Index.
Last verified . 29 shared benchmarks.
DeepSeek
NVIDIASummary
- They share 29 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 5 categories and Nemotron 3 Ultra in 3 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.0 to 34.4.
- The biggest single-benchmark swing is ProofBench: 56% for DeepSeek V4 Flash and 2% for Nemotron 3 Ultra.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $0.60 / $2.40 for Nemotron 3 Ultra.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 262K.
Side by side
| DeepSeek V4 Flash | Nemotron 3 Ultra | |
|---|---|---|
| Provider | DeepSeek | NVIDIA |
| Noometry Index | 53.6 | 44.7 |
| Released | 2026-04-24 | 2026-06-04 |
| Weights | Open | Open |
| Context window | 1M | 262K |
| Max output | 393K | 128K |
| Input $ / M tokens | $0.15 | $0.60 |
| Output $ / M tokens | $0.60 | $2.40 |
| Results tracked | 41 | 29 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Flash leads
DeepSeek V4 Flash: 47.6 (#56), Nemotron 3 Ultra: 37.7 (#174)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| FrontierCode | 18.8% | 13.6% |
| SciCode | 49.9% | 40.3% |
| WeirdML | 63% | 43.5% |
| LMArena Coding | 1457 | 1468 |
| LMArena WebDev | 1581 | — |
| ALE-Bench | 1,306 | — |
Reasoning DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.7 (#25), Nemotron 3 Ultra: 30.3 (#78)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| CritPt | 16.6% | 3.1% |
| Chess Puzzles | 33% | 12% |
| LMArena Hard Prompts | 1443 | 1452 |
| Mystery Game Puzzles | 34% | 20% |
| DTBench | 90.9% | 90.1% |
| LMCA | 41.7% | 36.9% |
| Epoch Capabilities Index | 154.51 | 146.2 |
| ARC-AGI-2 | 61.4% | — |
| SimpleBench | 61.1% | — |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 89.6% | — |
| ARC-AGI-1 | 89% | — |
Math DeepSeek V4 Flash leads
DeepSeek V4 Flash: 60.0 (#31), Nemotron 3 Ultra: 34.4 (#160)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 94.4% | 86.7% |
| ProofBench | 56% | 2% |
| LMArena Math | 1427 | 1456 |
| FrontierMath (Tiers 1-3) | 57.5% | — |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
Knowledge DeepSeek V4 Flash leads
DeepSeek V4 Flash: 55.5 (#43), Nemotron 3 Ultra: 52.5 (#56)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| GPQA Diamond | 91% | 85.4% |
| LMArena Expert | 1441 | 1472 |
| SimpleQA Verified | 33.6% | — |
Multilingual Too close to call
DeepSeek V4 Flash: 53.2 (#71), Nemotron 3 Ultra: 53.7 (#61)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| LMArena Non-English | 1420 | 1427 |
| LMArena Chinese | 1467 | 1498 |
| LMArena French | 1438 | 1475 |
| LMArena German | 1418 | 1471 |
| LMArena Korean | 1383 | 1387 |
| LMArena Russian | 1428 | 1415 |
| LMArena Spanish | 1437 | 1452 |
| LMArena Japanese | 1403 | — |
Instruction Following Too close to call
DeepSeek V4 Flash: 75.1 (#76), Nemotron 3 Ultra: 74.9 (#84)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| LMArena Instruction Following | 1421 | 1417 |
Long Context Too close to call
DeepSeek V4 Flash: 43.8 (#81), Nemotron 3 Ultra: 43.9 (#77)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| LMArena Longer Query | 1435 | 1437 |
Writing & Preference Nemotron 3 Ultra leads
DeepSeek V4 Flash: 64.0 (#59), Nemotron 3 Ultra: 66.3 (#35)
| Benchmark | DeepSeek V4 Flash | Nemotron 3 Ultra |
|---|---|---|
| LMArena Text | 1432 | 1444 |
| LMArena Creative Writing | 1403 | 1392 |
| EQ-Bench Creative Writing | 1559 | 1692 |
| LMArena Multi-Turn | 1449 | 1412 |
Frequently asked questions
Is DeepSeek V4 Flash better than Nemotron 3 Ultra?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 44.7 on the Noometry Index.
Which is cheaper, DeepSeek V4 Flash or Nemotron 3 Ultra?
DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; Nemotron 3 Ultra lists at $0.60 and $2.40.
Is DeepSeek V4 Flash or Nemotron 3 Ultra better for coding?
DeepSeek V4 Flash scores higher on coding benchmarks: 47.6 versus 37.7 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Flash does, with 1M tokens against 262K.
How many benchmarks do DeepSeek V4 Flash and Nemotron 3 Ultra share?
29 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and Nemotron 3 Ultra has 29.