Gemma 4 31B vs Nemotron 3 Ultra
Gemma 4 31B leads Nemotron 3 Ultra on coding, 39.0 to 30.0 on the Vibe Coding Index, and Gemma 4 31B costs 6.0x less per output token ($0.40 vs $2.40 per million). Updated October 23, 2026; 41 benchmarks have a result for both.
Gemma 4 31B or Nemotron 3 Ultra: which to pick
Pick Gemma 4 31B for coding: it scores 39.0 on the Vibe Coding Index against 30.0 for Nemotron 3 Ultra, though on 2 benchmarks against 3.
Pick Nemotron 3 Ultra for work measured by AA Coding Index or LMArena Text: coding, where it leads.
- Shared results
- 41
- Gemma 4 31B only
- 31
- Nemotron 3 Ultra only
- 30
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
Gemma 4 31B
Gemma 4 31B scores 39.0 against 30.0.
2 and 3 benchmarks behind the scores
Intelligence
AA Intelligence Index
Nemotron 3 Ultra
Nemotron 3 Ultra leads, 22.9 against 14.7.
Same board, both measured
Agentic work
AA Agentic Index
Nemotron 3 Ultra
Nemotron 3 Ultra leads, 20.1 against 4.2.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Nemotron 3 Ultra has a GPQA Diamond result, 86.1%.
Directional only
Knowledge
MMLU Pro
No call
Only Nemotron 3 Ultra has a MMLU Pro result, 85.8%.
Directional only
Web design
Design Arena: websites
No call
Only Nemotron 3 Ultra has a Design Arena: websites result, 1145.
Directional only
Long documents
Listed context window
No call
Both take 262K tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Gemma 4 31B
Gemma 4 31B $0.0003, Nemotron 3 Ultra $0.0018 per turn.
List prices
At a glance
- Coding index
- Gemma 4 31B 39.0 from 2 benchmarks, Nemotron 3 Ultra 30.0 from 3 benchmarks.
- Widest category
- Agentic work, 16.9 points apart on average over 2 shared benchmarks.
- Widest single gap
- τ²-Bench Telecom: Gemma 4 31B 65.5% against Nemotron 3 Ultra 83.3%, 17.8 points apart on a 0 to 100 scale.
- Output price
- Gemma 4 31B $0.40 against Nemotron 3 Ultra $2.40 per million output tokens, 6.0x apart.
- Context window
- Both take 262K tokens.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Gemma 4 31B | Bars | Nemotron 3 Ultra | Ties |
|---|---|---|---|---|
| Coding | 2 | 3 | 1 tie | |
| Intelligence | 0 | 2 | ||
| Agentic work | 0 | 2 | ||
| Reasoning | 1 | 5 | ||
| Math | 2 | 1 | ||
| Knowledge | 0 | 4 | ||
| General | 10 | 8 | ||
| All shared boards | 15 | 25 | 1 tie |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index23
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| LiveBench coding | 49.0% | Not measured | Only Gemma 4 31B measured |
| LMArena WebDev | 1368 | Not measured | Only Gemma 4 31B measured |
| AA Coding Index | 43.4 | 49.3 | Nemotron 3 Ultra |
| LiveCodeBench (Vals AI) | Not measured | 86.0% | Only Nemotron 3 Ultra measured |
| SWE-bench Verified (Vals AI) | Not measured | 69.0% | Only Nemotron 3 Ultra measured |
| Vibe Code Bench (Vals AI) | Not measured | 7.6% | Only Nemotron 3 Ultra measured |
| Code migration (Vals AI) | Not measured | 4.9% | Only Nemotron 3 Ultra measured |
| ProgramBench (Vals AI) | Not measured | 0.0% | Only Nemotron 3 Ultra measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 50.9% | Only Nemotron 3 Ultra measured |
| SciCode | 45.5% | 40.3% | Gemma 4 31B |
| Terminal-Bench Hard (Artificial Analysis run) | 36.4% | 36.4% | Tie |
| FrontierCode | Not measured | 13.6% | Only Nemotron 3 Ultra measured |
| WeirdML | 52.3% | 43.5% | Gemma 4 31B |
| ALE-Bench | 925.5 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: coding | 1459 | 1467 | Nemotron 3 Ultra |
| LMArena Text: software and IT | 1463 | 1468 | Nemotron 3 Ultra |
| LMArena WebDev: React | 1360 | Not measured | Only Gemma 4 31B measured |
| LMArena WebDev: HTML | 1381 | Not measured | Only Gemma 4 31B measured |
| SciCode | 8.9% | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 24 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| AA Intelligence Index | 14.7 | 22.9 | Nemotron 3 Ultra |
| Epoch Capabilities Index | 143.0 | 146.0 | Nemotron 3 Ultra |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| AA Agentic Index | 4.2 | 20.1 | Nemotron 3 Ultra |
| LiveBench research | 44.4% | Not measured | Only Gemma 4 31B measured |
| LiveBench browser | 4.0% | Not measured | Only Gemma 4 31B measured |
| LiveBench long agentic | 13.3% | Not measured | Only Gemma 4 31B measured |
| Finance Agent (Vals AI) | Not measured | 37.7% | Only Nemotron 3 Ultra measured |
| Legal Research Bench (Vals AI) | Not measured | 15.4% | Only Nemotron 3 Ultra measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 31.7% | Only Nemotron 3 Ultra measured |
| τ²-Bench Telecom | 65.5% | 83.3% | Nemotron 3 Ultra |
| APEX-Agents | Not measured | 22.7% | Only Nemotron 3 Ultra measured |
| MCP Atlas | Not measured | 63.1% | Only Nemotron 3 Ultra measured |
| Neither model has a result on the other 32 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index15
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 86.1% | Only Nemotron 3 Ultra measured |
| LiveBench reasoning | 68.3% | Not measured | Only Gemma 4 31B measured |
| GPQA Diamond (Artificial Analysis run) | 85.7% | 86.7% | Nemotron 3 Ultra |
| Humanity's Last Exam (Artificial Analysis run) | 23.6% | 28.4% | Nemotron 3 Ultra |
| GPQA Diamond (Epoch AI run) | 75.8% | 85.4% | Nemotron 3 Ultra |
| Mystery game puzzles | Not measured | 20.0% | Only Nemotron 3 Ultra measured |
| LMArena Text: hard prompts | 1448 | 1452 | Nemotron 3 Ultra |
| LMArena Text: expert | 1465 | 1472 | Nemotron 3 Ultra |
| LMArena Text: hard prompts in English | 1457 | 1453 | Gemma 4 31B |
| LMArena Vision: diagrams | 1283 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: homework | 1310 | Not measured | Only Gemma 4 31B measured |
| Kagi LLM Benchmark | 63.5% | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 11 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index21
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| OTIS Mock AIME | 73.3% | 86.7% | Nemotron 3 Ultra |
| ProofBench | Not measured | 2.0% | Only Nemotron 3 Ultra measured |
| LMArena Text: math | 1464 | 1457 | Gemma 4 31B |
| LMArena Text: mathematical work | 1465 | 1462 | Gemma 4 31B |
| Neither model has a result on the other 10 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 85.8% | Only Nemotron 3 Ultra measured |
| LegalBench (Vals AI) | Not measured | 82.1% | Only Nemotron 3 Ultra measured |
| CaseLaw (Vals AI) | 52.6% | Not measured | Only Gemma 4 31B measured |
| CorpFin (Vals AI) | Not measured | 65.5% | Only Nemotron 3 Ultra measured |
| TaxEval (Vals AI) | Not measured | 73.1% | Only Nemotron 3 Ultra measured |
| MortgageTax (Vals AI) | 61.4% | Not measured | Only Gemma 4 31B measured |
| MedCode (Vals AI) | Not measured | 38.6% | Only Nemotron 3 Ultra measured |
| SimpleQA Verified | 10.4% | Not measured | Only Gemma 4 31B measured |
| LMArena Text: science | 1457 | 1465 | Nemotron 3 Ultra |
| LMArena Text: medicine and healthcare | 1451 | 1455 | Nemotron 3 Ultra |
| LMArena Text: legal and government | 1438 | 1453 | Nemotron 3 Ultra |
| LMArena Text: business and finance | 1430 | 1435 | Nemotron 3 Ultra |
| FACTS Benchmark Suite | 45.3% | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1145 | Only Nemotron 3 Ultra measured |
| Design Arena: UI components | Not measured | 1135 | Only Nemotron 3 Ultra measured |
| Design Arena: game dev | Not measured | 1138 | Only Nemotron 3 Ultra measured |
| Design Arena: data viz | Not measured | 1142 | Only Nemotron 3 Ultra measured |
| Design Arena: 3D | Not measured | 1132 | Only Nemotron 3 Ultra measured |
| Design Arena: all code categories | Not measured | 1146 | Only Nemotron 3 Ultra measured |
| Design Arena: SVG | Not measured | 1069 | Only Nemotron 3 Ultra measured |
| Design Arena: ASCII art | Not measured | 1100 | Only Nemotron 3 Ultra measured |
| LiveBench design | 58.4% | Not measured | Only Gemma 4 31B measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index108
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| LiveBench data analysis | 53.3% | Not measured | Only Gemma 4 31B measured |
| SAGE (Vals AI) | 55.0% | Not measured | Only Gemma 4 31B measured |
| IFBench | 75.6% | 81.4% | Nemotron 3 Ultra |
| Long context reasoning (AA-LCR) | 69.7% | 79.3% | Nemotron 3 Ultra |
| DTBench | 82.7% | 90.1% | Nemotron 3 Ultra |
| LMCA | 39.3% | 36.9% | Gemma 4 31B |
| Surface Evolver Bench | 30.6% | Not measured | Only Gemma 4 31B measured |
| LMArena Text | 1443 | 1445 | Nemotron 3 Ultra |
| LMArena Text: instruction following | 1434 | 1418 | Gemma 4 31B |
| LMArena Text: creative writing | 1416 | 1398 | Gemma 4 31B |
| LMArena Vision | 1276 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: English | 1451 | 1454 | Nemotron 3 Ultra |
| LMArena Text: non-English | 1430 | 1427 | Gemma 4 31B |
| LMArena Text: Chinese | 1475 | 1498 | Nemotron 3 Ultra |
| LMArena Text: German | Not measured | 1474 | Only Nemotron 3 Ultra measured |
| LMArena Text: Korean | Not measured | 1385 | Only Nemotron 3 Ultra measured |
| LMArena Text: French | 1436 | 1473 | Nemotron 3 Ultra |
| LMArena Text: Spanish | 1442 | 1454 | Nemotron 3 Ultra |
| LMArena Text: Russian | 1461 | 1417 | Gemma 4 31B |
| LMArena Text: Polish | 1431 | 1385 | Gemma 4 31B |
| LMArena Text: multi-turn | 1452 | 1411 | Gemma 4 31B |
| LMArena Text: longer queries | 1446 | 1437 | Gemma 4 31B |
| LMArena Text: writing and language | 1431 | 1405 | Gemma 4 31B |
| LMArena Text: entertainment and media | 1405 | 1397 | Gemma 4 31B |
| LMArena Vision: English | 1271 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: Chinese | 1326 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: reading text | 1282 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: humor | 1280 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: captioning | 1218 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: recognizing things | 1226 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: creative writing | 1285 | Not measured | Only Gemma 4 31B measured |
| LMArena Document | 1425 | Not measured | Only Gemma 4 31B measured |
| EQ-Bench 4 | 1120 | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Gemma 4 31B | Nemotron 3 Ultra | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Gemma 4 31B costs less
Repository review
50K input + 3K output tokens
Gemma 4 31B costs less
Agent loop
200K input + 10K output tokens
Gemma 4 31B costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Google Gemma 4 31B | NVIDIA Nemotron 3 Ultra |
|---|---|---|
| Maker | NVIDIA | |
| Weights | Open weights | Open weights |
| Released | April 2, 2026 | June 4, 2026 |
| Context window | 262K | 262K |
| Input $/Mtok | $0.14 | $0.60 |
| Output $/Mtok | $0.40 | $2.40 |
| Coding index | 39.0 | 30.0 |
| Independent measurements behind it | 2 | 3 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Gemma 4 31B better than Nemotron 3 Ultra for coding?
Pick Gemma 4 31B for coding: it scores 39.0 on the Vibe Coding Index against 30.0 for Nemotron 3 Ultra, though on 2 benchmarks against 3. Pick Nemotron 3 Ultra for work measured by AA Coding Index or LMArena Text: coding, where it leads.
Which is cheaper, Gemma 4 31B or Nemotron 3 Ultra?
Gemma 4 31B costs $0.14 per million input tokens and $0.40 per million output tokens; Nemotron 3 Ultra costs $0.60 and $2.40. Cost per solved task is known for neither of them.
Which has the bigger context window, Gemma 4 31B or Nemotron 3 Ultra?
Neither: both take 262K tokens.
How many benchmarks is the Gemma 4 31B vs Nemotron 3 Ultra comparison based on?
41 benchmarks have a result for both models as of October 23, 2026. Gemma 4 31B's coding index rests on 2 independent benchmarks and Nemotron 3 Ultra's on 3. Where only one model has a result, the page says so instead of naming a winner.