Model comparison
DeepSeek-V3 vs Gemma 4 31B IT
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 39.5 on the Noometry Index.
Last verified . 25 shared benchmarks.
DeepSeek
GoogleSummary
- They share 25 benchmarks with published results for both. DeepSeek-V3 scores higher in 1 category and Gemma 4 31B IT in 7 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where Gemma 4 31B IT leads 43.2 to 32.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 73.3% for Gemma 4 31B IT.
- Gemma 4 31B IT is cheaper at $0.09 / $0.34 per million input/output tokens, against $0.24 / $0.90 for DeepSeek-V3.
- Gemma 4 31B IT accepts more context: 262K tokens versus 164K.
Side by side
| DeepSeek-V3 | Gemma 4 31B IT | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 39.5 | 43.5 |
| Released | 2024-12-26 | 2026-04-02 |
| Weights | Open | Open |
| Context window | 164K | 262K |
| Max output | 164K | 33K |
| Input $ / M tokens | $0.24 | $0.09 |
| Output $ / M tokens | $0.90 | $0.34 |
| Results tracked | 60 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek-V3: 42.3 (#106), Gemma 4 31B IT: 42.3 (#108)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| SciCode | 35.8% | 43.4% |
| WeirdML | 36.1% | 52.3% |
| LMArena Coding | 1368 | 1459 |
| Aider Polyglot | 55.1% | — |
| LMArena WebDev | — | 1366 |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 925.5 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Gemma 4 31B IT: —
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| METR Time Horizons | 49.6% | — |
Reasoning Gemma 4 31B IT leads
DeepSeek-V3: 20.5 (#236), Gemma 4 31B IT: 27.2 (#122)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 63.5% |
| CritPt | 0% | 1.4% |
| LMArena Hard Prompts | 1365 | 1448 |
| DTBench | 64.8% | 82.7% |
| LMCA | 15.5% | 39.3% |
| Epoch Capabilities Index | 135.94 | 142.74 |
| SimpleBench | 27.2% | — |
| NYT Connections (extended) | — | 70.6% |
| Chess Puzzles | — | 5% |
| Thematic Generalization | — | 53% |
| LiveBench Reasoning | 65.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| Surface Evolver Bench | — | 30.6% |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math Gemma 4 31B IT leads
DeepSeek-V3: 32.1 (#219), Gemma 4 31B IT: 43.2 (#81)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 73.3% |
| LMArena Math | 1373 | 1465 |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge Too close to call
DeepSeek-V3: 37.5 (#155), Gemma 4 31B IT: 37.9 (#151)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| GPQA Diamond | 67.6% | 75.8% |
| Vectara Hallucination Rate | 6.1% | 7.4% |
| LMArena Expert | 1351 | 1465 |
| SimpleQA Verified | — | 10.4% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, Gemma 4 31B IT: 41.6 (#34)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| LMArena Vision | — | 1277 |
| LMArena Document | — | 1425 |
Multilingual Gemma 4 31B IT leads
DeepSeek-V3: 48.5 (#143), Gemma 4 31B IT: 53.8 (#57)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| LMArena Non-English | 1358 | 1431 |
| LMArena Chinese | 1391 | 1476 |
| LMArena French | 1385 | 1435 |
| LMArena Russian | 1373 | 1460 |
| LMArena Spanish | 1358 | 1444 |
| LMArena German | 1374 | — |
| LMArena Japanese | 1333 | — |
| LMArena Korean | 1319 | — |
Instruction Following Gemma 4 31B IT leads
DeepSeek-V3: 72.8 (#130), Gemma 4 31B IT: 75.5 (#61)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| LMArena Instruction Following | 1345 | 1433 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
Long Context Gemma 4 31B IT leads
DeepSeek-V3: 34.0 (#253), Gemma 4 31B IT: 44.2 (#71)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| LMArena Longer Query | 1352 | 1446 |
| Fiction.LiveBench | 50% | — |
Writing & Preference Gemma 4 31B IT leads
DeepSeek-V3: 57.4 (#130), Gemma 4 31B IT: 60.5 (#96)
| Benchmark | DeepSeek-V3 | Gemma 4 31B IT |
|---|---|---|
| LMArena Text | 1375 | 1443 |
| LMArena Creative Writing | 1364 | 1415 |
| EQ-Bench Creative Writing | 1472 | 1368 |
| LMArena Multi-Turn | 1389 | 1452 |
| Short-Story Creative Writing | 77% | — |
| WildBench | 83% | — |
| EQ-Bench 4 | — | 1120 |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Gemma 4 31B IT?
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 39.5 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or Gemma 4 31B IT?
Gemma 4 31B IT is cheaper. It lists at $0.09 per million input tokens and $0.34 per million output tokens; DeepSeek-V3 lists at $0.24 and $0.90.
Is DeepSeek-V3 or Gemma 4 31B IT better for coding?
They score almost the same on coding (42.3 vs 42.3); test both on your own repository before choosing.
Which has the bigger context window?
Gemma 4 31B IT does, with 262K tokens against 164K.
How many benchmarks do DeepSeek-V3 and Gemma 4 31B IT share?
25 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Gemma 4 31B IT has 35.