Model comparison
GPT-3.5-turbo vs o1
o1 is the stronger model overall, scoring 40.6 to 23.1 on the Noometry Index. GPT-3.5-turbo costs 35× less per token, which makes it the better buy when o1's lead doesn't matter for your workload.
Last verified . 29 shared benchmarks.
Summary
- They share 29 benchmarks with published results for both. GPT-3.5-turbo scores higher in 0 categories and o1 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where o1 leads 41.5 to 10.0.
- The biggest single-benchmark swing is MATH Level 5: 15.9% for GPT-3.5-turbo and 94.7% for o1.
- GPT-3.5-turbo is cheaper at $0.50 / $1.50 per million input/output tokens, against $15 / $60 for o1.
- o1 accepts more context: 200K tokens versus 16K.
Side by side
| GPT-3.5-turbo | o1 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 23.1 | 40.6 |
| Released | 2023-03-01 | 2024-09-12 |
| Weights | Proprietary | Proprietary |
| Context window | 16K | 200K |
| Max output | 4K | 100K |
| Input $ / M tokens | $0.50 | $15 |
| Output $ / M tokens | $1.50 | $60 |
| Results tracked | 44 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o1 leads
GPT-3.5-turbo: 23.5 (#326), o1: 45.7 (#64)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| WeirdML | 3.5% | 47.6% |
| LMArena Coding | 1137 | 1367 |
| HumanEval+ | 70.7% | 89% |
| MBPP+ | 69.7% | 80.2% |
| Aider Polyglot | — | 61.7% |
| BigCodeBench Instruct | 39.1% | — |
| LiveBench Coding | — | 69.7% |
| BigCodeBench Complete | 50.6% | — |
| CadEval | — | 56% |
Agentic & Tool Use Not comparable
GPT-3.5-turbo: —, o1: 24.3 (#111)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| METR Time Horizons | 21.5% | 51.1% |
| Cybench | — | 10% |
Reasoning o1 leads
GPT-3.5-turbo: 13.3 (#328), o1: 27.5 (#103)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| Chess Puzzles | 0% | 15% |
| LMArena Hard Prompts | 1108 | 1372 |
| DTBench | 48.5% | 74.7% |
| LMCA | 9.7% | 22.3% |
| Epoch Capabilities Index | 118.48 | 141.92 |
| SimpleBench | — | 41.7% |
| ARC-AGI-1 | — | 30.7% |
| EnigmaEval | — | 5.7% |
| LiveBench Reasoning | — | 91.6% |
| Mystery Game Puzzles | 3% | — |
| LiveBench Data Analysis | — | 65.5% |
| Adversarial NLI | 58.1% | — |
| BIG-Bench Hard | 61.6% | — |
| CommonsenseQA 2.0 | 57% | — |
| ForecastBench | 50.4 | — |
| LiveBench | — | 75.7% |
| WinoGrande | 81.6% | — |
Math o1 leads
GPT-3.5-turbo: 6.0 (#321), o1: 34.9 (#150)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 0% | 14.7% |
| OTIS Mock AIME 2024-2025 | 2.2% | 73.3% |
| LMArena Math | 1142 | 1388 |
| MATH Level 5 | 15.9% | 94.7% |
| LiveBench Math | — | 80.3% |
| FrontierMath (Feb 2025 set) | — | 9.3% |
| GSM8K | 57.8% | — |
Knowledge o1 leads
GPT-3.5-turbo: 10.0 (#298), o1: 41.5 (#101)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| GPQA Diamond | 28% | 76.8% |
| LMArena Expert | 1070 | 1362 |
| Humanity's Last Exam | — | 8% |
| SimpleQA Verified | — | 41.1% |
| Confabulations | — | 11.7% |
| ARC (AI2) Challenge | 87.4% | — |
| BoolQ | 87% | — |
| MMLU | 71.4% | — |
| OpenBookQA | 86% | — |
| TriviaQA | 85.8% | — |
Multimodal Not comparable
GPT-3.5-turbo: —, o1: 33.6 (#87)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| LMArena Vision | — | 1168 |
| GeoBench | — | 80% |
| VPCT | — | 37% |
| SpatialViz-Bench | — | 41.4% |
Multilingual o1 leads
GPT-3.5-turbo: 31.6 (#253), o1: 48.8 (#137)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| LMArena Non-English | 1108 | 1359 |
| LMArena Chinese | 1075 | 1394 |
| LMArena French | 1117 | 1343 |
| LMArena German | 1090 | 1337 |
| LMArena Japanese | 1043 | 1346 |
| LMArena Korean | 1019 | 1396 |
| LMArena Russian | 1123 | 1356 |
| LMArena Spanish | 1120 | 1345 |
Instruction Following o1 leads
GPT-3.5-turbo: 58.1 (#257), o1: 75.0 (#82)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| LMArena Instruction Following | 1119 | 1368 |
| LiveBench Instruction Following | — | 81.5% |
Long Context o1 leads
GPT-3.5-turbo: 34.0 (#249), o1: 50.3 (#9)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| LMArena Longer Query | 1121 | 1378 |
| Fiction.LiveBench | — | 83.3% |
Writing & Preference o1 leads
GPT-3.5-turbo: 25.4 (#300), o1: 55.8 (#139)
| Benchmark | GPT-3.5-turbo | o1 |
|---|---|---|
| LMArena Text | 1125 | 1366 |
| LMArena Creative Writing | 1092 | 1347 |
| LMArena Multi-Turn | 1117 | 1368 |
| Short-Story Creative Writing | — | 70.2% |
| EQ-Bench Creative Writing | 451 | — |
| LiveBench Language | — | 65.4% |
Frequently asked questions
Is GPT-3.5-turbo better than o1?
o1 is the stronger model overall, scoring 40.6 to 23.1 on the Noometry Index. GPT-3.5-turbo costs 35× less per token, which makes it the better buy when o1's lead doesn't matter for your workload.
Which is cheaper, GPT-3.5-turbo or o1?
GPT-3.5-turbo is cheaper. It lists at $0.50 per million input tokens and $1.50 per million output tokens; o1 lists at $15 and $60.
Is GPT-3.5-turbo or o1 better for coding?
o1 scores higher on coding benchmarks: 45.7 versus 23.5 in the Noometry coding category.
Which has the bigger context window?
o1 does, with 200K tokens against 16K.
How many benchmarks do GPT-3.5-turbo and o1 share?
29 benchmarks have published results for both models. GPT-3.5-turbo has 44 scored results on Noometry and o1 has 52.