# Gemma 3 12B vs Gemma 3 4B

> Gemma 3 12B is the stronger model overall, scoring 31.8 to 27.9 on the Noometry Index. Gemma 3 4B costs 1.5× less per token, which makes it the better buy when Gemma 3 12B's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gemma-3-12b-vs-gemma-3-4b
- Last updated: 2026-10-10
- Shared benchmarks: 21

## Summary

- They share 21 benchmarks with published results for both. Gemma 3 12B scores higher in 8 categories and Gemma 3 4B in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Gemma 3 12B leads 26.5 to 11.8.
- The biggest single-benchmark swing is GPQA Diamond: 39.5% for Gemma 3 12B and 23.2% for Gemma 3 4B.
- Gemma 3 4B is cheaper at $0.04 / $0.08 per million input/output tokens, against $0.05 / $0.15 for Gemma 3 12B.

## Snapshot

| | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| Provider | Google | Google |
| Noometry Index | 31.8 | 27.9 |
| Rank | 255 | 320 |
| Context | 131K | 131K |
| Input $/M | $0.05 | $0.04 |
| Output $/M | $0.15 | $0.08 |
| Weights | Open | Open |

## Coding

- Gemma 3 12B: 31.2 (#276)
- Gemma 3 4B: 35.3 (#214)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| LMArena Coding | 1281 | 1230 |
| SciCode | 17.4% | — |

## Agentic & Tool Use

- Gemma 3 12B: 25.3 (#101)
- Gemma 3 4B: 20.7 (#134)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 30.4% | 19.6% |

## Reasoning

- Gemma 3 12B: 15.2 (#307)
- Gemma 3 4B: 12.7 (#330)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| Chess Puzzles | 0% | 0% |
| LMArena Hard Prompts | 1309 | 1253 |
| DTBench | 48.8% | 50.9% |
| LMCA | 4.5% | 2.8% |
| Epoch Capabilities Index | 123.49 | 116.02 |
| Kagi LLM Benchmark | — | 25.2% |
| CritPt | 0% | — |

## Math

- Gemma 3 12B: 21.0 (#273)
- Gemma 3 4B: 15.8 (#286)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 16.7% | 7.5% |
| LMArena Math | 1307 | 1239 |

## Knowledge

- Gemma 3 12B: 26.5 (#250)
- Gemma 3 4B: 11.8 (#294)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| GPQA Diamond | 39.5% | 23.2% |
| Vectara Hallucination Rate | 4.4% | 6.4% |
| LMArena Expert | 1248 | 1223 |

## Multimodal

- Gemma 3 12B: —
- Gemma 3 4B: —

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| MindCube | 46.7% | — |

## Multilingual

- Gemma 3 12B: 45.9 (#160)
- Gemma 3 4B: 42.7 (#189)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| LMArena Non-English | 1318 | 1273 |
| LMArena German | 1370 | 1281 |
| LMArena Russian | 1335 | 1294 |

## Instruction Following

- Gemma 3 12B: 68.8 (#179)
- Gemma 3 4B: 65.4 (#219)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| LMArena Instruction Following | 1299 | 1239 |

## Long Context

- Gemma 3 12B: 40.0 (#157)
- Gemma 3 4B: 38.6 (#189)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| LMArena Longer Query | 1317 | 1274 |

## Writing & Preference

- Gemma 3 12B: 47.7 (#205)
- Gemma 3 4B: 42.2 (#234)

| Benchmark | Gemma 3 12B | Gemma 3 4B |
|---|---|---|
| LMArena Text | 1334 | 1291 |
| LMArena Creative Writing | 1331 | 1270 |
| EQ-Bench Creative Writing | 1126 | 1068 |
| LMArena Multi-Turn | 1334 | 1255 |

## FAQ

### Is Gemma 3 12B better than Gemma 3 4B?

Gemma 3 12B is the stronger model overall, scoring 31.8 to 27.9 on the Noometry Index. Gemma 3 4B costs 1.5× less per token, which makes it the better buy when Gemma 3 12B's lead doesn't matter for your workload.

### Which is cheaper, Gemma 3 12B 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; Gemma 3 12B lists at $0.05 and $0.15.

### Is Gemma 3 12B or Gemma 3 4B better for coding?

Gemma 3 4B scores higher on coding benchmarks: 35.3 versus 31.2 in the Noometry coding category.

### Which has the bigger context window?

Both accept 131K tokens.

### How many benchmarks do Gemma 3 12B and Gemma 3 4B share?

21 benchmarks have published results for both models. Gemma 3 12B has 24 scored results on Noometry and Gemma 3 4B has 22.