Model comparison
Gemma 3 27B vs Gemma 4 31B IT
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 30.8 on the Noometry Index. Gemma 3 27B costs 1.5× less per token, which makes it the better buy when Gemma 4 31B IT's lead doesn't matter for your workload.
Last verified . 26 shared benchmarks.
GoogleSummary
- They share 26 benchmarks with published results for both. Gemma 3 27B scores higher in 0 categories and Gemma 4 31B IT in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in coding, where Gemma 4 31B IT leads 42.3 to 22.5.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 22.5% for Gemma 3 27B and 73.3% for Gemma 4 31B IT.
- Gemma 3 27B is cheaper at $0.08 / $0.16 per million input/output tokens, against $0.09 / $0.34 for Gemma 4 31B IT.
- Gemma 4 31B IT accepts more context: 262K tokens versus 131K.
Side by side
| Gemma 3 27B | Gemma 4 31B IT | |
|---|---|---|
| Provider | ||
| Noometry Index | 30.8 | 43.5 |
| Released | 2025-03-11 | 2026-04-02 |
| Weights | Open | Open |
| Context window | 131K | 262K |
| Max output | 8K | 33K |
| Input $ / M tokens | $0.08 | $0.09 |
| Output $ / M tokens | $0.16 | $0.34 |
| Results tracked | 43 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemma 4 31B IT leads
Gemma 3 27B: 22.5 (#334), Gemma 4 31B IT: 42.3 (#108)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| SciCode | 21.2% | 43.4% |
| LMArena Coding | 1322 | 1459 |
| Aider Polyglot | 4.9% | — |
| LMArena WebDev | — | 1366 |
| WeirdML | — | 52.3% |
| LiveBench Coding | 39.9% | — |
| ALE-Bench | — | 925.5 |
Agentic & Tool Use Not comparable
Gemma 3 27B: 25.1 (#110), Gemma 4 31B IT: —
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| Berkeley Function Calling Leaderboard | 29.5% | — |
Reasoning Gemma 4 31B IT leads
Gemma 3 27B: 16.7 (#301), Gemma 4 31B IT: 27.2 (#122)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| Kagi LLM Benchmark | 40.4% | 63.5% |
| CritPt | 0% | 1.4% |
| Chess Puzzles | 0% | 5% |
| LMArena Hard Prompts | 1340 | 1448 |
| DTBench | 52.5% | 82.7% |
| LMCA | 12.3% | 39.3% |
| Epoch Capabilities Index | 130.04 | 142.74 |
| NYT Connections (extended) | — | 70.6% |
| Thematic Generalization | — | 53% |
| LiveBench Reasoning | 43.8% | — |
| LiveBench Data Analysis | 51.5% | — |
| Surface Evolver Bench | — | 30.6% |
| LiveBench | 50% | — |
Math Gemma 4 31B IT leads
Gemma 3 27B: 25.9 (#265), Gemma 4 31B IT: 43.2 (#81)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 22.5% | 73.3% |
| LMArena Math | 1312 | 1465 |
| LiveBench Math | 55.4% | — |
| MATH Level 5 | 74% | — |
Knowledge Gemma 4 31B IT leads
Gemma 3 27B: 25.5 (#261), Gemma 4 31B IT: 37.9 (#151)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| GPQA Diamond | 47.7% | 75.8% |
| Vectara Hallucination Rate | 7.4% | 7.4% |
| LMArena Expert | 1304 | 1465 |
| SimpleQA Verified | — | 10.4% |
| Confabulations | 40.3% | — |
Multimodal Gemma 4 31B IT leads
Gemma 3 27B: 32.6 (#100), Gemma 4 31B IT: 41.6 (#34)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Vision | 1164 | 1277 |
| GeoBench | 52% | — |
| LMArena Document | — | 1425 |
Multilingual Gemma 4 31B IT leads
Gemma 3 27B: 46.9 (#155), Gemma 4 31B IT: 53.8 (#57)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Non-English | 1334 | 1431 |
| LMArena Chinese | 1346 | 1476 |
| LMArena French | 1368 | 1435 |
| LMArena Russian | 1349 | 1460 |
| LMArena Spanish | 1349 | 1444 |
| LMArena German | 1362 | — |
| LMArena Japanese | 1287 | — |
| LMArena Korean | 1308 | — |
Instruction Following Gemma 4 31B IT leads
Gemma 3 27B: 70.6 (#160), Gemma 4 31B IT: 75.5 (#61)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Instruction Following | 1321 | 1433 |
| LiveBench Instruction Following | 74.9% | — |
Long Context Gemma 4 31B IT leads
Gemma 3 27B: 27.6 (#293), Gemma 4 31B IT: 44.2 (#71)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Longer Query | 1333 | 1446 |
| Fiction.LiveBench | 33.3% | — |
Writing & Preference Gemma 4 31B IT leads
Gemma 3 27B: 52.5 (#168), Gemma 4 31B IT: 60.5 (#96)
| Benchmark | Gemma 3 27B | Gemma 4 31B IT |
|---|---|---|
| LMArena Text | 1358 | 1443 |
| LMArena Creative Writing | 1346 | 1415 |
| EQ-Bench Creative Writing | 1266 | 1368 |
| LMArena Multi-Turn | 1345 | 1452 |
| Short-Story Creative Writing | 79.9% | — |
| EQ-Bench 4 | — | 1120 |
| LiveBench Language | 34.6% | — |
Frequently asked questions
Is Gemma 3 27B better than Gemma 4 31B IT?
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 30.8 on the Noometry Index. Gemma 3 27B costs 1.5× less per token, which makes it the better buy when Gemma 4 31B IT's lead doesn't matter for your workload.
Which is cheaper, Gemma 3 27B or Gemma 4 31B IT?
Gemma 3 27B is cheaper. It lists at $0.08 per million input tokens and $0.16 per million output tokens; Gemma 4 31B IT lists at $0.09 and $0.34.
Is Gemma 3 27B or Gemma 4 31B IT better for coding?
Gemma 4 31B IT scores higher on coding benchmarks: 42.3 versus 22.5 in the Noometry coding category.
Which has the bigger context window?
Gemma 4 31B IT does, with 262K tokens against 131K.
How many benchmarks do Gemma 3 27B and Gemma 4 31B IT share?
26 benchmarks have published results for both models. Gemma 3 27B has 43 scored results on Noometry and Gemma 4 31B IT has 35.