Model comparison
Gemma 2 9B vs GPT-4o
GPT-4o is the stronger model overall, scoring 28.5 to 25.7 on the Noometry Index.
Last verified . 31 shared benchmarks.
GoogleSummary
- They share 31 benchmarks with published results for both. Gemma 2 9B scores higher in 2 categories and GPT-4o in 6 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where GPT-4o leads 52.8 to 32.3.
- The biggest single-benchmark swing is LiveBench Reasoning: 15.2% for Gemma 2 9B and 55.8% for GPT-4o.
- Gemma 2 9B has downloadable open weights; the other is API-only.
Side by side
| Gemma 2 9B | GPT-4o | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 25.7 | 28.5 |
| Released | 2024-06-24 | 2024-05-13 |
| Weights | Open | Proprietary |
| Context window | — | 128K |
| Max output | — | 16K |
| Input $ / M tokens | — | $2.50 |
| Output $ / M tokens | — | $10 |
| Results tracked | 35 | 72 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemma 2 9B leads
Gemma 2 9B: 29.0 (#299), GPT-4o: 25.1 (#323)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| BigCodeBench Instruct | 34.7% | 51.1% |
| LiveBench Coding | 22.5% | 51.4% |
| LMArena Coding | 1174 | 1297 |
| BigCodeBench Complete | 40.6% | 61.1% |
| SWE-bench Verified | — | 31% |
| SWE-bench Verified (bash only) | — | 21.6% |
| Aider Polyglot | — | 45.3% |
| GSO | — | 0% |
| WeirdML | — | 25.1% |
| CadEval | — | 26% |
| HumanEval+ | — | 87.2% |
| MBPP+ | — | 72.2% |
Agentic & Tool Use Not comparable
Gemma 2 9B: —, GPT-4o: 20.8 (#133)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| GDPval | — | 9.9% |
| TheAgentCompany | — | 8.6% |
| Cybench | — | 12.5% |
| BALROG | — | 32.3% |
| LMArena Search | — | 1006 |
| METR Time Horizons | — | 40.8% |
Reasoning Gemma 2 9B leads
Gemma 2 9B: 15.3 (#306), GPT-4o: 9.4 (#338)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| LiveBench Reasoning | 15.2% | 55.8% |
| LMArena Hard Prompts | 1172 | 1281 |
| LiveBench Data Analysis | 36.4% | 60.9% |
| Epoch Capabilities Index | 119.78 | 128.97 |
| LiveBench | 28.7% | 55.3% |
| ARC-AGI-2 | — | 0% |
| SimpleBench | — | 17.8% |
| ARC-AGI-1 | — | 4.5% |
| CritPt | — | 0% |
| Chess Puzzles | — | 13% |
| EnigmaEval | — | 0.8% |
| DTBench | — | 64.5% |
| LMCA | — | 16.6% |
| ForecastBench | — | 57.7 |
| PIQA | 83.7% | — |
Math Too close to call
Gemma 2 9B: 9.3 (#312), GPT-4o: 10.2 (#303)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 0.6% | 6.4% |
| LiveBench Math | 19.8% | 49.5% |
| LMArena Math | 1183 | 1285 |
| MATH Level 5 | 21% | 53.3% |
| FrontierMath (Tiers 1-3) | — | 0.4% |
| Omni-MATH | — | 29.3% |
| FrontierMath (Feb 2025 set) | — | 0.3% |
| GSM8K | 84.9% | — |
Knowledge GPT-4o leads
Gemma 2 9B: 9.7 (#300), GPT-4o: 28.8 (#235)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| GPQA Diamond | 27.5% | 49.2% |
| LMArena Expert | 1147 | 1250 |
| MMLU | 72.1% | 88.1% |
| Humanity's Last Exam | — | 2.7% |
| SimpleQA Verified | — | 26% |
| MMLU-Pro | — | 71.3% |
| Confabulations | — | 15.3% |
| Vectara Hallucination Rate | — | 9.6% |
| GPQA (HELM) | — | 52% |
| BoolQ | 85.7% | — |
Multimodal Not comparable
Gemma 2 9B: —, GPT-4o: 33.9 (#85)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| LMArena Vision | — | 1137 |
| Video-MME | — | 71.9% |
| GeoBench | — | 71% |
| VPCT | — | 40% |
| ScienceQA | — | 88.5% |
Multilingual GPT-4o leads
Gemma 2 9B: 36.8 (#233), GPT-4o: 43.4 (#181)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| LMArena Non-English | 1188 | 1283 |
| LMArena Chinese | 1185 | 1277 |
| LMArena French | 1189 | 1303 |
| LMArena German | 1186 | 1282 |
| LMArena Japanese | 1144 | 1257 |
| LMArena Korean | 1137 | 1234 |
| LMArena Russian | 1200 | 1286 |
| LMArena Spanish | 1200 | 1292 |
Instruction Following GPT-4o leads
Gemma 2 9B: 57.8 (#264), GPT-4o: 66.8 (#202)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| LiveBench Instruction Following | 52.6% | 68.6% |
| LMArena Instruction Following | 1178 | 1278 |
| IFEval | — | 81.7% |
Long Context GPT-4o leads
Gemma 2 9B: 36.3 (#228), GPT-4o: 39.4 (#174)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| LMArena Longer Query | 1197 | 1289 |
| Fiction.LiveBench | — | 66.7% |
Writing & Preference GPT-4o leads
Gemma 2 9B: 32.3 (#276), GPT-4o: 52.8 (#160)
| Benchmark | Gemma 2 9B | GPT-4o |
|---|---|---|
| LMArena Text | 1207 | 1300 |
| LMArena Creative Writing | 1206 | 1291 |
| LMArena Multi-Turn | 1193 | 1302 |
| LiveBench Language | 25.5% | 47.6% |
| Short-Story Creative Writing | — | 81.8% |
| EQ-Bench Creative Writing | 841 | — |
| WildBench | — | 82.8% |
Frequently asked questions
Is Gemma 2 9B better than GPT-4o?
GPT-4o is the stronger model overall, scoring 28.5 to 25.7 on the Noometry Index.
Is Gemma 2 9B or GPT-4o better for coding?
Gemma 2 9B scores higher on coding benchmarks: 29.0 versus 25.1 in the Noometry coding category.
How many benchmarks do Gemma 2 9B and GPT-4o share?
31 benchmarks have published results for both models. Gemma 2 9B has 35 scored results on Noometry and GPT-4o has 72.