Model comparison
DeepSeek-V3.2-Exp vs Nemotron 3 Ultra
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 42.5 on the Noometry Index.
Last verified . 28 shared benchmarks.
DeepSeek
NVIDIASummary
- They share 28 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 4 categories and Nemotron 3 Ultra in 5 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where DeepSeek-V3.2-Exp leads 32.7 to 23.4.
- The biggest single-benchmark swing is LMCA: 29.1% for DeepSeek-V3.2-Exp and 36.9% for Nemotron 3 Ultra.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.60 / $2.40 for Nemotron 3 Ultra.
- Nemotron 3 Ultra accepts more context: 262K tokens versus 164K.
Side by side
| DeepSeek-V3.2-Exp | Nemotron 3 Ultra | |
|---|---|---|
| Provider | DeepSeek | NVIDIA |
| Noometry Index | 44.3 | 42.5 |
| Released | 2025-09-29 | 2026-06-04 |
| Weights | Open | Open |
| Context window | 164K | 262K |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.26 | $0.60 |
| Output $ / M tokens | $0.38 | $2.40 |
| Results tracked | 49 | 30 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 46.5 (#65), Nemotron 3 Ultra: 38.1 (#182)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| SciCode | 38.9% | 40.3% |
| WeirdML | 39.5% | 43.5% |
| LMArena Coding | 1454 | 1468 |
| FrontierCode | — | 13.6% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| LMArena WebDev | 1362 | — |
| SWE-bench Multilingual | 59% | — |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 32.7 (#59), Nemotron 3 Ultra: 23.4 (#126)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| APEX-Agents | 21.3% | 22.7% |
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| Vending-Bench 2 | 1,034 | — |
Reasoning Nemotron 3 Ultra leads
DeepSeek-V3.2-Exp: 22.1 (#208), Nemotron 3 Ultra: 31.0 (#82)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| CritPt | 2.9% | 3.1% |
| Chess Puzzles | 14% | 12% |
| LMArena Hard Prompts | 1434 | 1452 |
| DTBench | 87.7% | 90.1% |
| LMCA | 29.1% | 36.9% |
| Epoch Capabilities Index | 146.27 | 146.17 |
| ARC-AGI-2 | 4% | — |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 36.7% | — |
| ARC-AGI-1 | 57% | — |
| Thematic Generalization | 65% | — |
| Mystery Game Puzzles | — | 20% |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 41.7 (#87), Nemotron 3 Ultra: 35.3 (#187)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 86.7% |
| ProofBench | 8% | 2% |
| LMArena Math | 1435 | 1457 |
| MathArena Final-Answer Competitions | 57.7% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Too close to call
DeepSeek-V3.2-Exp: 51.7 (#66), Nemotron 3 Ultra: 52.5 (#61)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| GPQA Diamond | 83.4% | 85.4% |
| LMArena Expert | 1436 | 1472 |
| Vectara Hallucination Rate | 5.3% | — |
Multilingual Nemotron 3 Ultra leads
DeepSeek-V3.2-Exp: 52.2 (#90), Nemotron 3 Ultra: 53.5 (#64)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| LMArena Non-English | 1409 | 1427 |
| LMArena Chinese | 1461 | 1497 |
| LMArena French | 1433 | 1472 |
| LMArena German | 1440 | 1471 |
| LMArena Korean | 1371 | 1386 |
| LMArena Russian | 1424 | 1417 |
| LMArena Spanish | 1440 | 1454 |
| LMArena Japanese | 1374 | — |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), Nemotron 3 Ultra: 74.7 (#88)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| LMArena Instruction Following | 1413 | 1418 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), Nemotron 3 Ultra: 43.9 (#81)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| LMArena Longer Query | 1428 | 1437 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference Nemotron 3 Ultra leads
DeepSeek-V3.2-Exp: 62.4 (#77), Nemotron 3 Ultra: 66.3 (#36)
| Benchmark | DeepSeek-V3.2-Exp | Nemotron 3 Ultra |
|---|---|---|
| LMArena Text | 1425 | 1445 |
| LMArena Creative Writing | 1403 | 1398 |
| EQ-Bench Creative Writing | 1515 | 1692 |
| LMArena Multi-Turn | 1427 | 1411 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Nemotron 3 Ultra?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 42.5 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or Nemotron 3 Ultra?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Nemotron 3 Ultra lists at $0.60 and $2.40.
Is DeepSeek-V3.2-Exp or Nemotron 3 Ultra better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.5 versus 38.1 in the Noometry coding category.
Which has the bigger context window?
Nemotron 3 Ultra does, with 262K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and Nemotron 3 Ultra share?
28 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Nemotron 3 Ultra has 30.