Model comparison
Gemma 2 27B vs Gemma 4 31B IT
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 29.4 on the Noometry Index.
Last verified . 19 shared benchmarks.
GoogleSummary
- They share 19 benchmarks with published results for both. Gemma 2 27B scores higher in 0 categories and Gemma 4 31B IT in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where Gemma 4 31B IT leads 43.2 to 10.7.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.4% for Gemma 2 27B and 73.3% for Gemma 4 31B IT.
- Gemma 4 31B IT is cheaper at $0.09 / $0.34 per million input/output tokens, against $0.65 / $0.65 for Gemma 2 27B.
- Gemma 4 31B IT accepts more context: 262K tokens versus 8K.
Side by side
| Gemma 2 27B | Gemma 4 31B IT | |
|---|---|---|
| Provider | ||
| Noometry Index | 29.4 | 43.5 |
| Released | 2024-06-24 | 2026-04-02 |
| Weights | Open | Open |
| Context window | 8K | 262K |
| Max output | 2K | 33K |
| Input $ / M tokens | $0.65 | $0.09 |
| Output $ / M tokens | $0.65 | $0.34 |
| Results tracked | 34 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemma 4 31B IT leads
Gemma 2 27B: 34.1 (#246), Gemma 4 31B IT: 42.3 (#108)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Coding | 1211 | 1459 |
| LMArena WebDev | — | 1366 |
| SciCode | — | 43.4% |
| WeirdML | — | 52.3% |
| BigCodeBench Instruct | 42.8% | — |
| LiveBench Coding | 36% | — |
| BigCodeBench Complete | 52.5% | — |
| ALE-Bench | — | 925.5 |
Reasoning Gemma 4 31B IT leads
Gemma 2 27B: 15.3 (#315), Gemma 4 31B IT: 27.2 (#122)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Hard Prompts | 1198 | 1448 |
| DTBench | 48% | 82.7% |
| LMCA | 7.1% | 39.3% |
| Epoch Capabilities Index | 122.08 | 142.74 |
| Kagi LLM Benchmark | — | 63.5% |
| NYT Connections (extended) | — | 70.6% |
| CritPt | — | 1.4% |
| Chess Puzzles | — | 5% |
| Thematic Generalization | — | 53% |
| LiveBench Reasoning | 28.1% | — |
| LiveBench Data Analysis | 47.9% | — |
| Surface Evolver Bench | — | 30.6% |
| LiveBench | 38.2% | — |
Math Gemma 4 31B IT leads
Gemma 2 27B: 10.7 (#311), Gemma 4 31B IT: 43.2 (#81)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.4% | 73.3% |
| LMArena Math | 1212 | 1465 |
| LiveBench Math | 26.5% | — |
| MATH Level 5 | 27.9% | — |
Knowledge Gemma 4 31B IT leads
Gemma 2 27B: 19.0 (#280), Gemma 4 31B IT: 37.9 (#151)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| GPQA Diamond | 36.5% | 75.8% |
| LMArena Expert | 1172 | 1465 |
| SimpleQA Verified | — | 10.4% |
| Confabulations | 27.1% | — |
| Vectara Hallucination Rate | — | 7.4% |
| MMLU | 75.7% | — |
Multimodal Not comparable
Gemma 2 27B: —, Gemma 4 31B IT: 41.6 (#34)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Vision | — | 1277 |
| LMArena Document | — | 1425 |
Multilingual Gemma 4 31B IT leads
Gemma 2 27B: 38.6 (#226), Gemma 4 31B IT: 53.8 (#57)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Non-English | 1217 | 1431 |
| LMArena Chinese | 1221 | 1476 |
| LMArena French | 1247 | 1435 |
| LMArena Russian | 1234 | 1460 |
| LMArena Spanish | 1228 | 1444 |
| LMArena German | 1209 | — |
| LMArena Japanese | 1175 | — |
| LMArena Korean | 1174 | — |
Instruction Following Gemma 4 31B IT leads
Gemma 2 27B: 60.5 (#249), Gemma 4 31B IT: 75.5 (#61)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Instruction Following | 1206 | 1433 |
| LiveBench Instruction Following | 58.1% | — |
Long Context Gemma 4 31B IT leads
Gemma 2 27B: 37.3 (#218), Gemma 4 31B IT: 44.2 (#71)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Longer Query | 1231 | 1446 |
Writing & Preference Gemma 4 31B IT leads
Gemma 2 27B: 44.2 (#225), Gemma 4 31B IT: 60.5 (#96)
| Benchmark | Gemma 2 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Text | 1231 | 1443 |
| LMArena Creative Writing | 1241 | 1415 |
| LMArena Multi-Turn | 1224 | 1452 |
| EQ-Bench Creative Writing | — | 1368 |
| EQ-Bench 4 | — | 1120 |
| LiveBench Language | 32.6% | — |
Frequently asked questions
Is Gemma 2 27B better than Gemma 4 31B IT?
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 29.4 on the Noometry Index.
Which is cheaper, Gemma 2 27B 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; Gemma 2 27B lists at $0.65 and $0.65.
Is Gemma 2 27B or Gemma 4 31B IT better for coding?
Gemma 4 31B IT scores higher on coding benchmarks: 42.3 versus 34.1 in the Noometry coding category.
Which has the bigger context window?
Gemma 4 31B IT does, with 262K tokens against 8K.
How many benchmarks do Gemma 2 27B and Gemma 4 31B IT share?
19 benchmarks have published results for both models. Gemma 2 27B has 34 scored results on Noometry and Gemma 4 31B IT has 35.