Model comparison
GPT-4o vs o3-mini
o3-mini is the stronger model overall, scoring 36.7 to 28.6 on the Noometry Index.
Last verified . 47 shared benchmarks.
Summary
- They share 47 benchmarks with published results for both. GPT-4o scores higher in 2 categories and o3-mini in 7 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where o3-mini leads 28.1 to 10.6.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 6.4% for GPT-4o and 76.9% for o3-mini.
- o3-mini is cheaper at $1.10 / $4.40 per million input/output tokens, against $2.50 / $10 for GPT-4o.
- o3-mini accepts more context: 200K tokens versus 128K.
Side by side
| GPT-4o | o3-mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 28.6 | 36.7 |
| Released | 2024-05-13 | 2024-12-20 |
| Weights | Proprietary | Proprietary |
| Context window | 128K | 200K |
| Max output | 16K | 100K |
| Input $ / M tokens | $2.50 | $1.10 |
| Output $ / M tokens | $10 | $4.40 |
| Results tracked | 72 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o3-mini leads
GPT-4o: 24.8 (#328), o3-mini: 40.8 (#132)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| Aider Polyglot | 45.3% | 60.4% |
| GSO | 0% | 1.3% |
| WeirdML | 25.1% | 43.7% |
| LiveBench Coding | 51.4% | 82.7% |
| LMArena Coding | 1297 | 1378 |
| CadEval | 26% | 54% |
| SWE-bench Verified | 31% | — |
| SWE-bench Verified (bash only) | 21.6% | — |
| SciCode | — | 39.8% |
| BigCodeBench Instruct | 51.1% | — |
| BigCodeBench Complete | 61.1% | — |
| HumanEval+ | 87.2% | — |
| MBPP+ | 72.2% | — |
Agentic & Tool Use o3-mini leads
GPT-4o: 21.0 (#141), o3-mini: 29.6 (#84)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| Cybench | 12.5% | 22.5% |
| GDPval | 9.9% | — |
| TheAgentCompany | 8.6% | — |
| BALROG | 32.3% | — |
| LMArena Search | 1006 | — |
| METR Time Horizons | 40.8% | — |
Reasoning o3-mini leads
GPT-4o: 9.4 (#343), o3-mini: 16.3 (#305)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| ARC-AGI-2 | 0% | 3% |
| SimpleBench | 17.8% | 22.8% |
| ARC-AGI-1 | 4.5% | 34.5% |
| CritPt | 0% | 0.3% |
| Chess Puzzles | 13% | 17% |
| LiveBench Reasoning | 55.8% | 89.6% |
| LMArena Hard Prompts | 1281 | 1366 |
| DTBench | 64.5% | 68.8% |
| LiveBench Data Analysis | 60.9% | 70.6% |
| LMCA | 16.6% | 19% |
| Epoch Capabilities Index | 128.97 | 140.34 |
| ForecastBench | 57.7 | 59.6 |
| LiveBench | 55.3% | 75.9% |
| EnigmaEval | 0.8% | — |
| Mystery Game Puzzles | — | 7% |
Math o3-mini leads
GPT-4o: 10.6 (#312), o3-mini: 28.1 (#244)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 0.4% | 18.6% |
| OTIS Mock AIME 2024-2025 | 6.4% | 76.9% |
| LiveBench Math | 49.5% | 77.3% |
| LMArena Math | 1285 | 1396 |
| MATH Level 5 | 53.3% | 96.5% |
| FrontierMath (Feb 2025 set) | 0.3% | 12.4% |
| FrontierMath Tier 4 | — | 0% |
| Omni-MATH | 29.3% | — |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge o3-mini leads
GPT-4o: 28.8 (#242), o3-mini: 38.3 (#146)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| GPQA Diamond | 49.2% | 77% |
| SimpleQA Verified | 26% | 15.3% |
| Confabulations | 15.3% | 17.9% |
| LMArena Expert | 1250 | 1364 |
| Humanity's Last Exam | 2.7% | — |
| MMLU-Pro | 71.3% | — |
| Vectara Hallucination Rate | 9.6% | — |
| GPQA (HELM) | 52% | — |
| MMLU | 88.1% | — |
Multimodal Not comparable
GPT-4o: 34.5 (#91), o3-mini: —
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| LMArena Vision | 1137 | — |
| Video-MME | 71.9% | — |
| GeoBench | 71% | — |
| VPCT | 40% | — |
| ScienceQA | 88.5% | — |
Multilingual o3-mini leads
GPT-4o: 43.2 (#186), o3-mini: 45.7 (#164)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| LMArena Non-English | 1283 | 1319 |
| LMArena Chinese | 1277 | 1379 |
| LMArena French | 1304 | 1334 |
| LMArena German | 1282 | 1303 |
| LMArena Japanese | 1257 | 1286 |
| LMArena Korean | 1234 | 1314 |
| LMArena Russian | 1286 | 1304 |
| LMArena Spanish | 1292 | 1321 |
Instruction Following o3-mini leads
GPT-4o: 66.6 (#207), o3-mini: 75.1 (#72)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| LiveBench Instruction Following | 68.6% | 84.4% |
| LMArena Instruction Following | 1278 | 1337 |
| IFEval | 81.7% | — |
Long Context GPT-4o leads
GPT-4o: 39.4 (#179), o3-mini: 33.8 (#256)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| Fiction.LiveBench | 66.7% | 50% |
| LMArena Longer Query | 1289 | 1343 |
Writing & Preference GPT-4o leads
GPT-4o: 52.6 (#166), o3-mini: 50.3 (#182)
| Benchmark | GPT-4o | o3-mini |
|---|---|---|
| LMArena Text | 1300 | 1337 |
| LMArena Creative Writing | 1292 | 1286 |
| Short-Story Creative Writing | 81.8% | 61.7% |
| LMArena Multi-Turn | 1302 | 1320 |
| LiveBench Language | 47.6% | 50.7% |
| WildBench | 82.8% | — |
Frequently asked questions
Is GPT-4o better than o3-mini?
o3-mini is the stronger model overall, scoring 36.7 to 28.6 on the Noometry Index.
Which is cheaper, GPT-4o or o3-mini?
o3-mini is cheaper. It lists at $1.10 per million input tokens and $4.40 per million output tokens; GPT-4o lists at $2.50 and $10.
Is GPT-4o or o3-mini better for coding?
o3-mini scores higher on coding benchmarks: 40.8 versus 24.8 in the Noometry coding category.
Which has the bigger context window?
o3-mini does, with 200K tokens against 128K.
How many benchmarks do GPT-4o and o3-mini share?
47 benchmarks have published results for both models. GPT-4o has 72 scored results on Noometry and o3-mini has 51.