Model comparison
Nemotron 3 Ultra vs Qwen2.5 72B Instruct
Nemotron 3 Ultra is the stronger model overall, scoring 42.5 to 31.9 on the Noometry Index.
Last verified . 22 shared benchmarks.
NVIDIA
Alibaba (Qwen)Summary
- They share 22 benchmarks with published results for both. Nemotron 3 Ultra scores higher in 9 categories and Qwen2.5 72B Instruct in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Nemotron 3 Ultra leads 52.5 to 27.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 86.7% for Nemotron 3 Ultra and 8.1% for Qwen2.5 72B Instruct.
- Nemotron 3 Ultra is cheaper at $0.50 / $2.20 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
- Nemotron 3 Ultra accepts more context: 262K tokens versus 131K.
Side by side
| Nemotron 3 Ultra | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | NVIDIA | Alibaba (Qwen) |
| Noometry Index | 42.5 | 31.9 |
| Released | 2026-06-04 | 2024-09 |
| Weights | Open | Open |
| Context window | 262K | 131K |
| Max output | 128K | 8K |
| Input $ / M tokens | $0.50 | $1.40 |
| Output $ / M tokens | $2.20 | $5.60 |
| Results tracked | 30 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Nemotron 3 Ultra leads
Nemotron 3 Ultra: 38.1 (#182), Qwen2.5 72B Instruct: 33.2 (#260)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| WeirdML | 43.5% | 16% |
| LMArena Coding | 1468 | 1292 |
| FrontierCode | 13.6% | — |
| SciCode | 40.3% | — |
| BigCodeBench Instruct | — | 45.8% |
| BigCodeBench Complete | — | 55.9% |
Agentic & Tool Use Nemotron 3 Ultra leads
Nemotron 3 Ultra: 23.4 (#126), Qwen2.5 72B Instruct: 22.1 (#133)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| APEX-Agents | 22.7% | — |
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| METR Time Horizons | — | 35.8% |
Reasoning Nemotron 3 Ultra leads
Nemotron 3 Ultra: 31.0 (#82), Qwen2.5 72B Instruct: 22.3 (#199)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1452 | 1271 |
| DTBench | 90.1% | 62.9% |
| LMCA | 36.9% | 13.4% |
| Epoch Capabilities Index | 146.17 | 129 |
| CritPt | 3.1% | — |
| Chess Puzzles | 12% | — |
| Mystery Game Puzzles | 20% | — |
| BIG-Bench Hard | — | 79.8% |
| ForecastBench | — | 57.5 |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math Nemotron 3 Ultra leads
Nemotron 3 Ultra: 35.3 (#187), Qwen2.5 72B Instruct: 19.3 (#287)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 86.7% | 8.1% |
| LMArena Math | 1457 | 1283 |
| ProofBench | 2% | — |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |
Knowledge Nemotron 3 Ultra leads
Nemotron 3 Ultra: 52.5 (#61), Qwen2.5 72B Instruct: 27.0 (#253)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 85.4% | 49.1% |
| LMArena Expert | 1472 | 1245 |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |
Multilingual Nemotron 3 Ultra leads
Nemotron 3 Ultra: 53.5 (#64), Qwen2.5 72B Instruct: 41.0 (#213)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1427 | 1252 |
| LMArena Chinese | 1497 | 1272 |
| LMArena French | 1472 | 1280 |
| LMArena German | 1471 | 1234 |
| LMArena Korean | 1386 | 1188 |
| LMArena Russian | 1417 | 1264 |
| LMArena Spanish | 1454 | 1256 |
| LMArena Japanese | — | 1180 |
Instruction Following Nemotron 3 Ultra leads
Nemotron 3 Ultra: 74.7 (#88), Qwen2.5 72B Instruct: 65.5 (#221)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1418 | 1254 |
| IFEval | — | 80.6% |
Long Context Nemotron 3 Ultra leads
Nemotron 3 Ultra: 43.9 (#81), Qwen2.5 72B Instruct: 38.9 (#188)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1437 | 1282 |
Writing & Preference Nemotron 3 Ultra leads
Nemotron 3 Ultra: 66.3 (#36), Qwen2.5 72B Instruct: 46.7 (#215)
| Benchmark | Nemotron 3 Ultra | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1445 | 1269 |
| LMArena Creative Writing | 1398 | 1221 |
| LMArena Multi-Turn | 1411 | 1272 |
| EQ-Bench Creative Writing | 1692 | — |
| WildBench | — | 80.2% |
Frequently asked questions
Is Nemotron 3 Ultra better than Qwen2.5 72B Instruct?
Nemotron 3 Ultra is the stronger model overall, scoring 42.5 to 31.9 on the Noometry Index.
Which is cheaper, Nemotron 3 Ultra or Qwen2.5 72B Instruct?
Nemotron 3 Ultra is cheaper. It lists at $0.50 per million input tokens and $2.20 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is Nemotron 3 Ultra or Qwen2.5 72B Instruct better for coding?
Nemotron 3 Ultra scores higher on coding benchmarks: 38.1 versus 33.2 in the Noometry coding category.
Which has the bigger context window?
Nemotron 3 Ultra does, with 262K tokens against 131K.
How many benchmarks do Nemotron 3 Ultra and Qwen2.5 72B Instruct share?
22 benchmarks have published results for both models. Nemotron 3 Ultra has 30 scored results on Noometry and Qwen2.5 72B Instruct has 43.