Model comparison
Nemotron 3 Ultra vs Qwen3 Max
Qwen3 Max is the stronger model overall, scoring 43.7 to 42.5 on the Noometry Index. Nemotron 3 Ultra costs 2.6× less per token, which makes it the better buy when Qwen3 Max's lead doesn't matter for your workload.
Last verified . 23 shared benchmarks.
NVIDIA
Alibaba (Qwen)Summary
- They share 23 benchmarks with published results for both. Nemotron 3 Ultra scores higher in 4 categories and Qwen3 Max in 4 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Nemotron 3 Ultra leads 31.0 to 22.6.
- The biggest single-benchmark swing is Mystery Game Puzzles: 20% for Nemotron 3 Ultra and 5% for Qwen3 Max.
- Nemotron 3 Ultra is cheaper at $0.50 / $2.20 per million input/output tokens, against $1.20 / $6 for Qwen3 Max.
- Nemotron 3 Ultra has downloadable open weights; the other is API-only.
Side by side
| Nemotron 3 Ultra | Qwen3 Max | |
|---|---|---|
| Provider | NVIDIA | Alibaba (Qwen) |
| Noometry Index | 42.5 | 43.7 |
| Released | 2026-06-04 | 2025-09-23 |
| Weights | Open | Proprietary |
| Context window | 262K | 262K |
| Max output | 128K | 66K |
| Input $ / M tokens | $0.50 | $1.20 |
| Output $ / M tokens | $2.20 | $6 |
| Results tracked | 30 | 33 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3 Max leads
Nemotron 3 Ultra: 38.1 (#182), Qwen3 Max: 43.0 (#93)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| LMArena Coding | 1468 | 1456 |
| FrontierCode | 13.6% | — |
| SciCode | 40.3% | — |
| WeirdML | 43.5% | — |
| ALE-Bench | — | 370.45 |
Agentic & Tool Use Not comparable
Nemotron 3 Ultra: 23.4 (#126), Qwen3 Max: —
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| APEX-Agents | 22.7% | — |
| Vending-Bench 2 | — | 71.56 |
Reasoning Nemotron 3 Ultra leads
Nemotron 3 Ultra: 31.0 (#82), Qwen3 Max: 22.6 (#190)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| Chess Puzzles | 12% | 4% |
| LMArena Hard Prompts | 1452 | 1448 |
| Mystery Game Puzzles | 20% | 5% |
| DTBench | 90.1% | 82.1% |
| LMCA | 36.9% | 28.3% |
| Epoch Capabilities Index | 146.17 | 142.38 |
| Kagi LLM Benchmark | — | 72.5% |
| NYT Connections (extended) | — | 30.1% |
| CritPt | 3.1% | — |
Math Qwen3 Max leads
Nemotron 3 Ultra: 35.3 (#187), Qwen3 Max: 38.7 (#131)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 86.7% | 73.3% |
| LMArena Math | 1457 | 1446 |
| FrontierMath (Tiers 1-3) | — | 18.9% |
| ProofBench | 2% | — |
| MATH Level 5 | — | 97.1% |
Knowledge Nemotron 3 Ultra leads
Nemotron 3 Ultra: 52.5 (#61), Qwen3 Max: 48.1 (#78)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| GPQA Diamond | 85.4% | 72.6% |
| LMArena Expert | 1472 | 1455 |
| SimpleQA Verified | — | 48.7% |
Multilingual Too close to call
Nemotron 3 Ultra: 53.5 (#64), Qwen3 Max: 53.7 (#62)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| LMArena Non-English | 1427 | 1429 |
| LMArena Chinese | 1497 | 1478 |
| LMArena French | 1472 | 1449 |
| LMArena German | 1471 | 1463 |
| LMArena Korean | 1386 | 1399 |
| LMArena Russian | 1417 | 1428 |
| LMArena Spanish | 1454 | 1462 |
| LMArena Japanese | — | 1397 |
Instruction Following Too close to call
Nemotron 3 Ultra: 74.7 (#88), Qwen3 Max: 74.8 (#87)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| LMArena Instruction Following | 1418 | 1419 |
Long Context Nemotron 3 Ultra leads
Nemotron 3 Ultra: 43.9 (#81), Qwen3 Max: 41.6 (#134)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| LMArena Longer Query | 1437 | 1438 |
| Fiction.LiveBench | — | 66.7% |
| CL-bench | — | 14.5% |
Writing & Preference Nemotron 3 Ultra leads
Nemotron 3 Ultra: 66.3 (#36), Qwen3 Max: 62.4 (#76)
| Benchmark | Nemotron 3 Ultra | Qwen3 Max |
|---|---|---|
| LMArena Text | 1445 | 1439 |
| LMArena Creative Writing | 1398 | 1402 |
| LMArena Multi-Turn | 1411 | 1446 |
| EQ-Bench Creative Writing | 1692 | — |
Frequently asked questions
Is Nemotron 3 Ultra better than Qwen3 Max?
Qwen3 Max is the stronger model overall, scoring 43.7 to 42.5 on the Noometry Index. Nemotron 3 Ultra costs 2.6× less per token, which makes it the better buy when Qwen3 Max's lead doesn't matter for your workload.
Which is cheaper, Nemotron 3 Ultra or Qwen3 Max?
Nemotron 3 Ultra is cheaper. It lists at $0.50 per million input tokens and $2.20 per million output tokens; Qwen3 Max lists at $1.20 and $6.
Is Nemotron 3 Ultra or Qwen3 Max better for coding?
Qwen3 Max scores higher on coding benchmarks: 43.0 versus 38.1 in the Noometry coding category.
Which has the bigger context window?
Both accept 262K tokens.
How many benchmarks do Nemotron 3 Ultra and Qwen3 Max share?
23 benchmarks have published results for both models. Nemotron 3 Ultra has 30 scored results on Noometry and Qwen3 Max has 33.