Model comparison
Gemma 2 27B vs o1
o1 is the stronger model overall, scoring 40.6 to 29.2 on the Noometry Index. Gemma 2 27B costs 40× less per token, which makes it the better buy when o1's lead doesn't matter for your workload.
Last verified . 31 shared benchmarks.
GoogleSummary
- They share 31 benchmarks with published results for both. Gemma 2 27B scores higher in 0 categories and o1 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where o1 leads 34.9 to 10.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.4% for Gemma 2 27B and 73.3% for o1.
- Gemma 2 27B is cheaper at $0.65 / $0.65 per million input/output tokens, against $15 / $60 for o1.
- o1 accepts more context: 200K tokens versus 8K.
- Gemma 2 27B has downloadable open weights; the other is API-only.
Side by side
| Gemma 2 27B | o1 | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 29.2 | 40.6 |
| Released | 2024-06-24 | 2024-09-12 |
| Weights | Open | Proprietary |
| Context window | 8K | 200K |
| Max output | 2K | 100K |
| Input $ / M tokens | $0.65 | $15 |
| Output $ / M tokens | $0.65 | $60 |
| Results tracked | 34 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o1 leads
Gemma 2 27B: 33.5 (#243), o1: 45.7 (#64)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| LiveBench Coding | 36% | 69.7% |
| LMArena Coding | 1211 | 1367 |
| Aider Polyglot | — | 61.7% |
| WeirdML | — | 47.6% |
| BigCodeBench Instruct | 42.8% | — |
| BigCodeBench Complete | 52.5% | — |
| CadEval | — | 56% |
| HumanEval+ | — | 89% |
| MBPP+ | — | 80.2% |
Agentic & Tool Use Not comparable
Gemma 2 27B: —, o1: 24.3 (#111)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| Cybench | — | 10% |
| METR Time Horizons | — | 51.1% |
Reasoning o1 leads
Gemma 2 27B: 14.8 (#313), o1: 27.5 (#103)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| LiveBench Reasoning | 28.1% | 91.6% |
| LMArena Hard Prompts | 1199 | 1372 |
| DTBench | 48% | 74.7% |
| LiveBench Data Analysis | 47.9% | 65.5% |
| LMCA | 7.1% | 22.3% |
| Epoch Capabilities Index | 122.06 | 141.92 |
| LiveBench | 38.2% | 75.7% |
| SimpleBench | — | 41.7% |
| ARC-AGI-1 | — | 30.7% |
| Chess Puzzles | — | 15% |
| EnigmaEval | — | 5.7% |
Math o1 leads
Gemma 2 27B: 10.0 (#306), o1: 34.9 (#150)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.4% | 73.3% |
| LiveBench Math | 26.5% | 80.3% |
| LMArena Math | 1212 | 1388 |
| MATH Level 5 | 27.9% | 94.7% |
| FrontierMath (Tiers 1-3) | — | 14.7% |
| FrontierMath (Feb 2025 set) | — | 9.3% |
Knowledge o1 leads
Gemma 2 27B: 18.9 (#275), o1: 41.5 (#101)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| GPQA Diamond | 36.5% | 76.8% |
| Confabulations | 27.1% | 11.7% |
| LMArena Expert | 1172 | 1362 |
| Humanity's Last Exam | — | 8% |
| SimpleQA Verified | — | 41.1% |
| MMLU | 75.7% | — |
Multimodal Not comparable
Gemma 2 27B: —, o1: 33.6 (#87)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| LMArena Vision | — | 1168 |
| GeoBench | — | 80% |
| VPCT | — | 37% |
| SpatialViz-Bench | — | 41.4% |
Multilingual o1 leads
Gemma 2 27B: 38.8 (#221), o1: 48.8 (#137)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| LMArena Non-English | 1217 | 1359 |
| LMArena Chinese | 1221 | 1394 |
| LMArena French | 1246 | 1343 |
| LMArena German | 1209 | 1337 |
| LMArena Japanese | 1175 | 1346 |
| LMArena Korean | 1174 | 1396 |
| LMArena Russian | 1234 | 1356 |
| LMArena Spanish | 1228 | 1345 |
Instruction Following o1 leads
Gemma 2 27B: 60.7 (#244), o1: 75.0 (#82)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| LiveBench Instruction Following | 58.1% | 81.5% |
| LMArena Instruction Following | 1206 | 1368 |
Long Context o1 leads
Gemma 2 27B: 37.3 (#213), o1: 50.3 (#9)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| LMArena Longer Query | 1231 | 1378 |
| Fiction.LiveBench | — | 83.3% |
Writing & Preference o1 leads
Gemma 2 27B: 44.4 (#220), o1: 55.8 (#139)
| Benchmark | Gemma 2 27B | o1 |
|---|---|---|
| LMArena Text | 1231 | 1366 |
| LMArena Creative Writing | 1241 | 1347 |
| LMArena Multi-Turn | 1224 | 1368 |
| LiveBench Language | 32.6% | 65.4% |
| Short-Story Creative Writing | — | 70.2% |
Frequently asked questions
Is Gemma 2 27B better than o1?
o1 is the stronger model overall, scoring 40.6 to 29.2 on the Noometry Index. Gemma 2 27B costs 40× less per token, which makes it the better buy when o1's lead doesn't matter for your workload.
Which is cheaper, Gemma 2 27B or o1?
Gemma 2 27B is cheaper. It lists at $0.65 per million input tokens and $0.65 per million output tokens; o1 lists at $15 and $60.
Is Gemma 2 27B or o1 better for coding?
o1 scores higher on coding benchmarks: 45.7 versus 33.5 in the Noometry coding category.
Which has the bigger context window?
o1 does, with 200K tokens against 8K.
How many benchmarks do Gemma 2 27B and o1 share?
31 benchmarks have published results for both models. Gemma 2 27B has 34 scored results on Noometry and o1 has 52.