Model comparison
DeepSeek V4 Pro vs Gemma 4 31B IT
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.5 on the Noometry Index. Gemma 4 31B IT costs 6.5× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.
Last verified . 32 shared benchmarks.
DeepSeek
GoogleSummary
- They share 32 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 8 categories and Gemma 4 31B IT in 0 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 27.2.
- The biggest single-benchmark swing is SimpleQA Verified: 52.9% for DeepSeek V4 Pro and 10.4% for Gemma 4 31B IT.
- Gemma 4 31B IT is cheaper at $0.09 / $0.34 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 262K.
Side by side
| DeepSeek V4 Pro | Gemma 4 31B IT | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 54.3 | 43.5 |
| Released | 2026-04-24 | 2026-04-02 |
| Weights | Open | Open |
| Context window | 1M | 262K |
| Max output | 393K | 33K |
| Input $ / M tokens | $0.66 | $0.09 |
| Output $ / M tokens | $1.98 | $0.34 |
| Results tracked | 48 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), Gemma 4 31B IT: 42.3 (#108)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| LMArena WebDev | 1582 | 1366 |
| SciCode | 51% | 43.4% |
| WeirdML | 66.2% | 52.3% |
| LMArena Coding | 1470 | 1459 |
| ALE-Bench | 1,403 | 925.5 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
Agentic & Tool Use Not comparable
DeepSeek V4 Pro: 32.8 (#58), Gemma 4 31B IT: —
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| APEX-Agents | 47.3% | — |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), Gemma 4 31B IT: 27.2 (#122)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| Kagi LLM Benchmark | 53.5% | 63.5% |
| NYT Connections (extended) | 91.3% | 70.6% |
| CritPt | 18% | 1.4% |
| Chess Puzzles | 47% | 5% |
| LMArena Hard Prompts | 1461 | 1448 |
| DTBench | 93.9% | 82.7% |
| LMCA | 45.5% | 39.3% |
| Surface Evolver Bench | 40% | 30.6% |
| Epoch Capabilities Index | 155.31 | 142.74 |
| ARC-AGI-2 | 61.3% | — |
| ARC-AGI-1 | 90.5% | — |
| Thematic Generalization | — | 53% |
| Mystery Game Puzzles | 43% | — |
| ForecastBench | 56.1 | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), Gemma 4 31B IT: 43.2 (#81)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 73.3% |
| LMArena Math | 1455 | 1465 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), Gemma 4 31B IT: 37.9 (#151)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| GPQA Diamond | 91.7% | 75.8% |
| SimpleQA Verified | 52.9% | 10.4% |
| Vectara Hallucination Rate | 8.6% | 7.4% |
| LMArena Expert | 1464 | 1465 |
Multimodal Not comparable
DeepSeek V4 Pro: —, Gemma 4 31B IT: 41.6 (#34)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| LMArena Vision | — | 1277 |
| LMArena Document | — | 1425 |
Multilingual Too close to call
DeepSeek V4 Pro: 54.4 (#45), Gemma 4 31B IT: 53.8 (#57)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| LMArena Non-English | 1439 | 1431 |
| LMArena Chinese | 1486 | 1476 |
| LMArena French | 1472 | 1435 |
| LMArena Russian | 1453 | 1460 |
| LMArena Spanish | 1458 | 1444 |
| LMArena German | 1458 | — |
| LMArena Japanese | 1445 | — |
| LMArena Korean | 1447 | — |
Instruction Following Too close to call
DeepSeek V4 Pro: 76.1 (#47), Gemma 4 31B IT: 75.5 (#61)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| LMArena Instruction Following | 1448 | 1433 |
Long Context Too close to call
DeepSeek V4 Pro: 45.0 (#51), Gemma 4 31B IT: 44.2 (#71)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| LMArena Longer Query | 1458 | 1446 |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), Gemma 4 31B IT: 60.5 (#96)
| Benchmark | DeepSeek V4 Pro | Gemma 4 31B IT |
|---|---|---|
| LMArena Text | 1451 | 1443 |
| LMArena Creative Writing | 1446 | 1415 |
| EQ-Bench Creative Writing | 1553 | 1368 |
| EQ-Bench 4 | 1166 | 1120 |
| LMArena Multi-Turn | 1467 | 1452 |
Frequently asked questions
Is DeepSeek V4 Pro better than Gemma 4 31B IT?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.5 on the Noometry Index. Gemma 4 31B IT costs 6.5× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.
Which is cheaper, DeepSeek V4 Pro 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 V4 Pro lists at $0.66 and $1.98.
Is DeepSeek V4 Pro or Gemma 4 31B IT better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 42.3 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Pro does, with 1M tokens against 262K.
How many benchmarks do DeepSeek V4 Pro and Gemma 4 31B IT share?
32 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Gemma 4 31B IT has 35.