Model comparison
Gemma 2 27B vs Gemma 3 4B
Gemma 2 27B is the stronger model overall, scoring 29.4 to 28.1 on the Noometry Index. Gemma 3 4B costs 13× less per token, which makes it the better buy when Gemma 2 27B's lead doesn't matter for your workload.
Last verified . 17 shared benchmarks.
GoogleSummary
- They share 17 benchmarks with published results for both. Gemma 2 27B scores higher in 3 categories and Gemma 3 4B in 5 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Gemma 2 27B leads 19.0 to 11.8.
- The biggest single-benchmark swing is GPQA Diamond: 36.5% for Gemma 2 27B and 23.2% for Gemma 3 4B.
- Gemma 3 4B is cheaper at $0.04 / $0.08 per million input/output tokens, against $0.65 / $0.65 for Gemma 2 27B.
- Gemma 3 4B accepts more context: 131K tokens versus 8K.
Side by side
| Gemma 2 27B | Gemma 3 4B | |
|---|---|---|
| Provider | ||
| Noometry Index | 29.4 | 28.1 |
| Released | 2024-06-24 | 2025-03-12 |
| Weights | Open | Open |
| Context window | 8K | 131K |
| Max output | 2K | 4K |
| Input $ / M tokens | $0.65 | $0.04 |
| Output $ / M tokens | $0.65 | $0.08 |
| Results tracked | 34 | 22 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemma 3 4B leads
Gemma 2 27B: 34.1 (#246), Gemma 3 4B: 35.9 (#215)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| LMArena Coding | 1211 | 1230 |
| BigCodeBench Instruct | 42.8% | — |
| LiveBench Coding | 36% | — |
| BigCodeBench Complete | 52.5% | — |
Agentic & Tool Use Not comparable
Gemma 2 27B: —, Gemma 3 4B: 20.9 (#142)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 19.6% |
Reasoning Gemma 2 27B leads
Gemma 2 27B: 15.3 (#315), Gemma 3 4B: 13.2 (#335)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| LMArena Hard Prompts | 1198 | 1253 |
| DTBench | 48% | 50.9% |
| LMCA | 7.1% | 2.8% |
| Epoch Capabilities Index | 122.08 | 116.02 |
| Kagi LLM Benchmark | — | 25.2% |
| Chess Puzzles | — | 0% |
| LiveBench Reasoning | 28.1% | — |
| LiveBench Data Analysis | 47.9% | — |
| LiveBench | 38.2% | — |
Math Gemma 3 4B leads
Gemma 2 27B: 10.7 (#311), Gemma 3 4B: 16.8 (#292)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.4% | 7.5% |
| LMArena Math | 1212 | 1239 |
| LiveBench Math | 26.5% | — |
| MATH Level 5 | 27.9% | — |
Knowledge Gemma 2 27B leads
Gemma 2 27B: 19.0 (#280), Gemma 3 4B: 11.8 (#299)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| GPQA Diamond | 36.5% | 23.2% |
| LMArena Expert | 1172 | 1223 |
| Confabulations | 27.1% | — |
| Vectara Hallucination Rate | — | 6.4% |
| MMLU | 75.7% | — |
Multilingual Gemma 3 4B leads
Gemma 2 27B: 38.6 (#226), Gemma 3 4B: 42.5 (#194)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| LMArena Non-English | 1217 | 1273 |
| LMArena German | 1209 | 1281 |
| LMArena Russian | 1234 | 1294 |
| LMArena Chinese | 1221 | — |
| LMArena French | 1247 | — |
| LMArena Japanese | 1175 | — |
| LMArena Korean | 1174 | — |
| LMArena Spanish | 1228 | — |
Instruction Following Gemma 3 4B leads
Gemma 2 27B: 60.5 (#249), Gemma 3 4B: 65.2 (#225)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| LMArena Instruction Following | 1206 | 1239 |
| LiveBench Instruction Following | 58.1% | — |
Long Context Gemma 3 4B leads
Gemma 2 27B: 37.3 (#218), Gemma 3 4B: 38.7 (#194)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| LMArena Longer Query | 1231 | 1273 |
Writing & Preference Gemma 2 27B leads
Gemma 2 27B: 44.2 (#225), Gemma 3 4B: 42.0 (#239)
| Benchmark | Gemma 2 27B | Gemma 3 4B |
|---|---|---|
| LMArena Text | 1231 | 1291 |
| LMArena Creative Writing | 1241 | 1271 |
| LMArena Multi-Turn | 1224 | 1255 |
| EQ-Bench Creative Writing | — | 1068 |
| LiveBench Language | 32.6% | — |
Frequently asked questions
Is Gemma 2 27B better than Gemma 3 4B?
Gemma 2 27B is the stronger model overall, scoring 29.4 to 28.1 on the Noometry Index. Gemma 3 4B costs 13× less per token, which makes it the better buy when Gemma 2 27B's lead doesn't matter for your workload.
Which is cheaper, Gemma 2 27B 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 2 27B lists at $0.65 and $0.65.
Is Gemma 2 27B or Gemma 3 4B better for coding?
Gemma 3 4B scores higher on coding benchmarks: 35.9 versus 34.1 in the Noometry coding category.
Which has the bigger context window?
Gemma 3 4B does, with 131K tokens against 8K.
How many benchmarks do Gemma 2 27B and Gemma 3 4B share?
17 benchmarks have published results for both models. Gemma 2 27B has 34 scored results on Noometry and Gemma 3 4B has 22.