# DeepSeek-V2.5 (Sep 2024) vs Gemma 3 4B

> DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 27.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v2-5-vs-gemma-3-4b
- Last updated: 2026-10-10
- Shared benchmarks: 12

## Summary

- They share 12 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 6 categories and Gemma 3 4B in 2 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V2.5 (Sep 2024) leads 34.2 to 11.8.

## Snapshot

| | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| Provider | DeepSeek | Google |
| Noometry Index | 37.0 | 27.9 |
| Rank | 197 | 320 |
| Context | — | 131K |
| Input $/M | — | $0.04 |
| Output $/M | — | $0.08 |
| Weights | Open | Open |

## Coding

- DeepSeek-V2.5 (Sep 2024): 31.3 (#274)
- Gemma 3 4B: 35.3 (#214)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Coding | 1310 | 1230 |
| Aider Polyglot | 17.8% | — |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |

## Agentic & Tool Use

- DeepSeek-V2.5 (Sep 2024): —
- Gemma 3 4B: 20.7 (#134)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 19.6% |

## Reasoning

- DeepSeek-V2.5 (Sep 2024): 24.5 (#145)
- Gemma 3 4B: 12.7 (#330)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1253 |
| Kagi LLM Benchmark | — | 25.2% |
| Chess Puzzles | — | 0% |
| DTBench | — | 50.9% |
| LMCA | — | 2.8% |
| Epoch Capabilities Index | — | 116.02 |

## Math

- DeepSeek-V2.5 (Sep 2024): 33.3 (#177)
- Gemma 3 4B: 15.8 (#286)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Math | 1289 | 1239 |
| OTIS Mock AIME 2024-2025 | — | 7.5% |

## Knowledge

- DeepSeek-V2.5 (Sep 2024): 34.2 (#190)
- Gemma 3 4B: 11.8 (#294)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Expert | 1266 | 1223 |
| GPQA Diamond | — | 23.2% |
| Vectara Hallucination Rate | — | 6.4% |

## Multilingual

- DeepSeek-V2.5 (Sep 2024): 42.7 (#188)
- Gemma 3 4B: 42.7 (#189)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Non-English | 1273 | 1273 |
| LMArena German | 1258 | 1281 |
| LMArena Russian | 1289 | 1294 |
| LMArena Chinese | 1317 | — |
| LMArena French | 1288 | — |
| LMArena Japanese | 1228 | — |
| LMArena Korean | 1209 | — |
| LMArena Spanish | 1248 | — |

## Instruction Following

- DeepSeek-V2.5 (Sep 2024): 67.7 (#189)
- Gemma 3 4B: 65.4 (#219)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Instruction Following | 1280 | 1239 |

## Long Context

- DeepSeek-V2.5 (Sep 2024): 39.5 (#169)
- Gemma 3 4B: 38.6 (#189)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Longer Query | 1301 | 1274 |

## Writing & Preference

- DeepSeek-V2.5 (Sep 2024): 50.0 (#182)
- Gemma 3 4B: 42.2 (#234)

| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 4B |
|---|---|---|
| LMArena Text | 1294 | 1291 |
| LMArena Creative Writing | 1285 | 1270 |
| LMArena Multi-Turn | 1297 | 1255 |
| EQ-Bench Creative Writing | — | 1068 |

## FAQ

### Is DeepSeek-V2.5 (Sep 2024) better than Gemma 3 4B?

DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 27.9 on the Noometry Index.

### Is DeepSeek-V2.5 (Sep 2024) or Gemma 3 4B better for coding?

Gemma 3 4B scores higher on coding benchmarks: 35.3 versus 31.3 in the Noometry coding category.

### How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Gemma 3 4B share?

12 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Gemma 3 4B has 22.