Model comparison
Gemma 2 27B vs GPT-4 Turbo
GPT-4 Turbo is the stronger model overall, scoring 30.3 to 29.2 on the Noometry Index. Gemma 2 27B costs 23× less per token, which makes it the better buy when GPT-4 Turbo's lead doesn't matter for your workload.
Last verified . 27 shared benchmarks.
GoogleSummary
- They share 27 benchmarks with published results for both. Gemma 2 27B scores higher in 2 categories and GPT-4 Turbo in 6 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where GPT-4 Turbo leads 24.3 to 18.9.
- The biggest single-benchmark swing is MATH Level 5: 27.9% for Gemma 2 27B and 46.7% for GPT-4 Turbo.
- Gemma 2 27B is cheaper at $0.65 / $0.65 per million input/output tokens, against $10 / $30 for GPT-4 Turbo.
- GPT-4 Turbo accepts more context: 128K tokens versus 8K.
- Gemma 2 27B has downloadable open weights; the other is API-only.
Side by side
| Gemma 2 27B | GPT-4 Turbo | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 29.2 | 30.3 |
| Released | 2024-06-24 | 2023-11-06 |
| Weights | Open | Proprietary |
| Context window | 8K | 128K |
| Max output | 2K | 4K |
| Input $ / M tokens | $0.65 | $10 |
| Output $ / M tokens | $0.65 | $30 |
| Results tracked | 34 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Gemma 2 27B: 33.5 (#243), GPT-4 Turbo: 33.4 (#247)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| BigCodeBench Instruct | 42.8% | 48.2% |
| LMArena Coding | 1211 | 1268 |
| BigCodeBench Complete | 52.5% | 58.2% |
| WeirdML | — | 18% |
| LiveBench Coding | 36% | — |
| HumanEval+ | — | 86.6% |
| MBPP+ | — | 73.3% |
Agentic & Tool Use Not comparable
Gemma 2 27B: —, GPT-4 Turbo: —
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| METR Time Horizons | — | 36.7% |
Reasoning Too close to call
Gemma 2 27B: 14.8 (#313), GPT-4 Turbo: 14.9 (#311)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| LMArena Hard Prompts | 1199 | 1251 |
| DTBench | 48% | 61.6% |
| LMCA | 7.1% | 9.8% |
| Epoch Capabilities Index | 122.06 | 127.25 |
| SimpleBench | — | 25.1% |
| Chess Puzzles | — | 6% |
| LiveBench Reasoning | 28.1% | — |
| LiveBench Data Analysis | 47.9% | — |
| ForecastBench | — | 59.4 |
| LiveBench | 38.2% | — |
Math Gemma 2 27B leads
Gemma 2 27B: 10.0 (#306), GPT-4 Turbo: 8.6 (#316)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.4% | 6.7% |
| LMArena Math | 1212 | 1272 |
| MATH Level 5 | 27.9% | 46.7% |
| FrontierMath (Tiers 1-3) | — | 0.7% |
| LiveBench Math | 26.5% | — |
Knowledge GPT-4 Turbo leads
Gemma 2 27B: 18.9 (#275), GPT-4 Turbo: 24.3 (#263)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| GPQA Diamond | 36.5% | 46.6% |
| Confabulations | 27.1% | 28.4% |
| LMArena Expert | 1172 | 1223 |
| MMLU | 75.7% | 81.3% |
Multimodal Not comparable
Gemma 2 27B: —, GPT-4 Turbo: 30.0 (#104)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| LMArena Vision | — | 1090 |
Multilingual GPT-4 Turbo leads
Gemma 2 27B: 38.8 (#221), GPT-4 Turbo: 40.7 (#211)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| LMArena Non-English | 1217 | 1245 |
| LMArena Chinese | 1221 | 1242 |
| LMArena French | 1246 | 1275 |
| LMArena German | 1209 | 1259 |
| LMArena Japanese | 1175 | 1194 |
| LMArena Korean | 1174 | 1187 |
| LMArena Russian | 1234 | 1259 |
| LMArena Spanish | 1228 | 1260 |
Instruction Following GPT-4 Turbo leads
Gemma 2 27B: 60.7 (#244), GPT-4 Turbo: 66.0 (#211)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| LMArena Instruction Following | 1206 | 1250 |
| LiveBench Instruction Following | 58.1% | — |
Long Context Too close to call
Gemma 2 27B: 37.3 (#213), GPT-4 Turbo: 38.0 (#200)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| LMArena Longer Query | 1231 | 1254 |
Writing & Preference GPT-4 Turbo leads
Gemma 2 27B: 44.4 (#220), GPT-4 Turbo: 47.9 (#201)
| Benchmark | Gemma 2 27B | GPT-4 Turbo |
|---|---|---|
| LMArena Text | 1231 | 1272 |
| LMArena Creative Writing | 1241 | 1269 |
| LMArena Multi-Turn | 1224 | 1267 |
| LiveBench Language | 32.6% | — |
Frequently asked questions
Is Gemma 2 27B better than GPT-4 Turbo?
GPT-4 Turbo is the stronger model overall, scoring 30.3 to 29.2 on the Noometry Index. Gemma 2 27B costs 23× less per token, which makes it the better buy when GPT-4 Turbo's lead doesn't matter for your workload.
Which is cheaper, Gemma 2 27B or GPT-4 Turbo?
Gemma 2 27B is cheaper. It lists at $0.65 per million input tokens and $0.65 per million output tokens; GPT-4 Turbo lists at $10 and $30.
Is Gemma 2 27B or GPT-4 Turbo better for coding?
They score almost the same on coding (33.5 vs 33.4); test both on your own repository before choosing.
Which has the bigger context window?
GPT-4 Turbo does, with 128K tokens against 8K.
How many benchmarks do Gemma 2 27B and GPT-4 Turbo share?
27 benchmarks have published results for both models. Gemma 2 27B has 34 scored results on Noometry and GPT-4 Turbo has 36.