# DeepSeek V4 Flash vs Gemma 3 4B

> DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 28.1 on the Noometry Index. Gemma 3 4B costs 5.2× less per token, which makes it the better buy when DeepSeek V4 Flash's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/deepseek-v4-flash-vs-gemma-3-4b
- Last updated: 2026-10-12
- Shared benchmarks: 20

## Summary

- They share 20 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 8 categories and Gemma 3 4B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek V4 Flash leads 55.4 to 11.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 94.4% for DeepSeek V4 Flash and 7.5% for Gemma 3 4B.
- Gemma 3 4B is cheaper at $0.04 / $0.08 per million input/output tokens, against $0.15 / $0.60 for DeepSeek V4 Flash.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 131K.

## Snapshot

| | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| Provider | DeepSeek | Google |
| Noometry Index | 53.6 | 28.1 |
| Rank | 35 | 326 |
| Context | 1M | 131K |
| Input $/M | $0.15 | $0.04 |
| Output $/M | $0.60 | $0.08 |
| Weights | Open | Open |

## Coding

- DeepSeek V4 Flash: 47.9 (#59)
- Gemma 3 4B: 35.9 (#215)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| LMArena Coding | 1457 | 1230 |
| FrontierCode | 18.8% | — |
| LMArena WebDev | 1582 | — |
| SciCode | 49.9% | — |
| WeirdML | 63% | — |
| ALE-Bench | 1,306 | — |

## Agentic & Tool Use

- DeepSeek V4 Flash: —
- Gemma 3 4B: 20.9 (#142)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 19.6% |

## Reasoning

- DeepSeek V4 Flash: 53.7 (#30)
- Gemma 3 4B: 13.2 (#335)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| Kagi LLM Benchmark | 52.2% | 25.2% |
| Chess Puzzles | 33% | 0% |
| LMArena Hard Prompts | 1444 | 1253 |
| DTBench | 90.9% | 50.9% |
| LMCA | 41.7% | 2.8% |
| Epoch Capabilities Index | 154.49 | 116.02 |
| ARC-AGI-2 | 61.4% | — |
| SimpleBench | 61.1% | — |
| NYT Connections (extended) | 89.6% | — |
| ARC-AGI-1 | 89% | — |
| CritPt | 16.6% | — |
| Mystery Game Puzzles | 34% | — |

## Math

- DeepSeek V4 Flash: 60.3 (#37)
- Gemma 3 4B: 16.8 (#292)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 94.4% | 7.5% |
| LMArena Math | 1427 | 1239 |
| FrontierMath (Tiers 1-3) | 57.5% | — |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
| ProofBench | 56% | — |

## Knowledge

- DeepSeek V4 Flash: 55.4 (#48)
- Gemma 3 4B: 11.8 (#299)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| GPQA Diamond | 91% | 23.2% |
| LMArena Expert | 1441 | 1223 |
| SimpleQA Verified | 33.6% | — |
| Vectara Hallucination Rate | — | 6.4% |

## Multilingual

- DeepSeek V4 Flash: 53.0 (#72)
- Gemma 3 4B: 42.5 (#194)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| LMArena Non-English | 1420 | 1273 |
| LMArena German | 1418 | 1281 |
| LMArena Russian | 1428 | 1294 |
| LMArena Chinese | 1468 | — |
| LMArena French | 1439 | — |
| LMArena Japanese | 1406 | — |
| LMArena Korean | 1384 | — |
| LMArena Spanish | 1436 | — |

## Instruction Following

- DeepSeek V4 Flash: 74.9 (#81)
- Gemma 3 4B: 65.2 (#225)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| LMArena Instruction Following | 1421 | 1239 |

## Long Context

- DeepSeek V4 Flash: 43.8 (#85)
- Gemma 3 4B: 38.7 (#194)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| LMArena Longer Query | 1434 | 1273 |

## Writing & Preference

- DeepSeek V4 Flash: 63.8 (#61)
- Gemma 3 4B: 42.0 (#239)

| Benchmark | DeepSeek V4 Flash | Gemma 3 4B |
|---|---|---|
| LMArena Text | 1432 | 1291 |
| LMArena Creative Writing | 1403 | 1271 |
| EQ-Bench Creative Writing | 1559 | 1068 |
| LMArena Multi-Turn | 1449 | 1255 |

## FAQ

### Is DeepSeek V4 Flash better than Gemma 3 4B?

DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 28.1 on the Noometry Index. Gemma 3 4B costs 5.2× less per token, which makes it the better buy when DeepSeek V4 Flash's lead doesn't matter for your workload.

### Which is cheaper, DeepSeek V4 Flash or Gemma 3 4B?

Gemma 3 4B is cheaper. It lists at $0.04 per million input tokens and $0.08 per million output tokens; DeepSeek V4 Flash lists at $0.15 and $0.60.

### Is DeepSeek V4 Flash or Gemma 3 4B better for coding?

DeepSeek V4 Flash scores higher on coding benchmarks: 47.9 versus 35.9 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Flash does, with 1M tokens against 131K.

### How many benchmarks do DeepSeek V4 Flash and Gemma 3 4B share?

20 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and Gemma 3 4B has 22.