Model comparison
Gemma 3 27B vs GPT-4.1
GPT-4.1 is the stronger model overall, scoring 35.7 to 30.5 on the Noometry Index. Gemma 3 27B costs 35× less per token, which makes it the better buy when GPT-4.1's lead doesn't matter for your workload.
Last verified . 32 shared benchmarks.
GoogleSummary
- They share 32 benchmarks with published results for both. Gemma 3 27B scores higher in 2 categories and GPT-4.1 in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-4.1 leads 40.0 to 27.6.
- The biggest single-benchmark swing is Aider Polyglot: 4.9% for Gemma 3 27B and 52.4% for GPT-4.1.
- Gemma 3 27B is cheaper at $0.08 / $0.16 per million input/output tokens, against $2 / $8 for GPT-4.1.
- GPT-4.1 accepts more context: 1.05M tokens versus 131K.
- Gemma 3 27B has downloadable open weights; the other is API-only.
Side by side
| Gemma 3 27B | GPT-4.1 | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 30.5 | 35.7 |
| Released | 2025-03-11 | 2025-04-14 |
| Weights | Open | Proprietary |
| Context window | 131K | 1.05M |
| Max output | 8K | 33K |
| Input $ / M tokens | $0.08 | $2 |
| Output $ / M tokens | $0.16 | $8 |
| Results tracked | 43 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-4.1 leads
Gemma 3 27B: 22.1 (#329), GPT-4.1: 34.2 (#228)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| Aider Polyglot | 4.9% | 52.4% |
| LMArena Coding | 1322 | 1391 |
| SWE-bench Verified | — | 48.5% |
| SWE-bench Verified (bash only) | — | 39.6% |
| SciCode | 21.2% | — |
| WeirdML | — | 39% |
| LiveBench Coding | 39.9% | — |
| CadEval | — | 42% |
| ALE-Bench | — | 558.1 |
Agentic & Tool Use GPT-4.1 leads
Gemma 3 27B: 24.9 (#105), GPT-4.1: 34.4 (#40)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| Berkeley Function Calling Leaderboard | 29.5% | 54% |
Reasoning Gemma 3 27B leads
Gemma 3 27B: 16.3 (#297), GPT-4.1: 11.7 (#334)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| Kagi LLM Benchmark | 40.4% | 52.3% |
| Chess Puzzles | 0% | 6% |
| LMArena Hard Prompts | 1340 | 1384 |
| DTBench | 52.5% | 68.3% |
| LMCA | 12.3% | 25.6% |
| Epoch Capabilities Index | 130.04 | 136.79 |
| ARC-AGI-2 | — | 0.4% |
| SimpleBench | — | 27% |
| ARC-AGI-1 | — | 5.5% |
| CritPt | 0% | — |
| EnigmaEval | — | 2.2% |
| LiveBench Reasoning | 43.8% | — |
| LiveBench Data Analysis | 51.5% | — |
| ForecastBench | — | 61.5 |
| LiveBench | 50% | — |
Math Gemma 3 27B leads
Gemma 3 27B: 24.4 (#261), GPT-4.1: 21.4 (#269)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 22.5% | 38.3% |
| LMArena Math | 1312 | 1369 |
| MATH Level 5 | 74% | 83% |
| FrontierMath (Tiers 1-3) | — | 6% |
| Omni-MATH | — | 47.1% |
| LiveBench Math | 55.4% | — |
| FrontierMath (Feb 2025 set) | — | 5.5% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge GPT-4.1 leads
Gemma 3 27B: 25.4 (#256), GPT-4.1: 37.0 (#152)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| GPQA Diamond | 47.7% | 66.9% |
| Vectara Hallucination Rate | 7.4% | 5.6% |
| LMArena Expert | 1303 | 1364 |
| Humanity's Last Exam | — | 5.4% |
| SimpleQA Verified | — | 31.1% |
| MMLU-Pro | — | 81.1% |
| Confabulations | 40.3% | — |
| GPQA (HELM) | — | 65.9% |
Multimodal GPT-4.1 leads
Gemma 3 27B: 32.0 (#94), GPT-4.1: 37.5 (#66)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| LMArena Vision | 1165 | 1210 |
| GeoBench | 52% | 72% |
Multilingual GPT-4.1 leads
Gemma 3 27B: 47.0 (#150), GPT-4.1: 49.6 (#128)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| LMArena Non-English | 1334 | 1370 |
| LMArena Chinese | 1347 | 1382 |
| LMArena French | 1368 | 1383 |
| LMArena German | 1362 | 1380 |
| LMArena Japanese | 1287 | 1320 |
| LMArena Korean | 1308 | 1339 |
| LMArena Russian | 1349 | 1377 |
| LMArena Spanish | 1348 | 1375 |
Instruction Following Too close to call
Gemma 3 27B: 70.8 (#155), GPT-4.1: 71.5 (#148)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| LMArena Instruction Following | 1321 | 1367 |
| LiveBench Instruction Following | 74.9% | — |
| IFEval | — | 83.8% |
Long Context GPT-4.1 leads
Gemma 3 27B: 27.6 (#288), GPT-4.1: 40.0 (#158)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| Fiction.LiveBench | 33.3% | 63.9% |
| LMArena Longer Query | 1333 | 1385 |
Writing & Preference GPT-4.1 leads
Gemma 3 27B: 52.7 (#161), GPT-4.1: 57.8 (#120)
| Benchmark | Gemma 3 27B | GPT-4.1 |
|---|---|---|
| LMArena Text | 1358 | 1383 |
| LMArena Creative Writing | 1346 | 1363 |
| EQ-Bench Creative Writing | 1266 | 1420 |
| LMArena Multi-Turn | 1344 | 1398 |
| Short-Story Creative Writing | 79.9% | — |
| WildBench | — | 85.4% |
| LiveBench Language | 34.6% | — |
Frequently asked questions
Is Gemma 3 27B better than GPT-4.1?
GPT-4.1 is the stronger model overall, scoring 35.7 to 30.5 on the Noometry Index. Gemma 3 27B costs 35× less per token, which makes it the better buy when GPT-4.1's lead doesn't matter for your workload.
Which is cheaper, Gemma 3 27B or GPT-4.1?
Gemma 3 27B is cheaper. It lists at $0.08 per million input tokens and $0.16 per million output tokens; GPT-4.1 lists at $2 and $8.
Is Gemma 3 27B or GPT-4.1 better for coding?
GPT-4.1 scores higher on coding benchmarks: 34.2 versus 22.1 in the Noometry coding category.
Which has the bigger context window?
GPT-4.1 does, with 1.05M tokens against 131K.
How many benchmarks do Gemma 3 27B and GPT-4.1 share?
32 benchmarks have published results for both models. Gemma 3 27B has 43 scored results on Noometry and GPT-4.1 has 52.