Model comparison
GPT-4o vs GPT-4o mini
GPT-4o is the stronger model overall, scoring 28.6 to 25.5 on the Noometry Index. GPT-4o mini costs 17× less per token, which makes it the better buy when GPT-4o's lead doesn't matter for your workload.
Last verified . 54 shared benchmarks.
Summary
- They share 54 benchmarks with published results for both. GPT-4o scores higher in 9 categories and GPT-4o mini in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where GPT-4o leads 52.6 to 39.5.
- The biggest single-benchmark swing is Aider Polyglot: 45.3% for GPT-4o and 3.6% for GPT-4o mini.
- GPT-4o mini is cheaper at $0.15 / $0.60 per million input/output tokens, against $2.50 / $10 for GPT-4o.
Side by side
| GPT-4o | GPT-4o mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 28.6 | 25.5 |
| Released | 2024-05-13 | 2024-07-18 |
| Weights | Proprietary | Proprietary |
| Context window | 128K | 128K |
| Max output | 16K | 16K |
| Input $ / M tokens | $2.50 | $0.15 |
| Output $ / M tokens | $10 | $0.60 |
| Results tracked | 72 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-4o leads
GPT-4o: 24.8 (#328), GPT-4o mini: 22.0 (#335)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| Aider Polyglot | 45.3% | 3.6% |
| WeirdML | 25.1% | 11.8% |
| BigCodeBench Instruct | 51.1% | 46.1% |
| LiveBench Coding | 51.4% | 43.1% |
| LMArena Coding | 1297 | 1290 |
| BigCodeBench Complete | 61.1% | 57.4% |
| HumanEval+ | 87.2% | 83.5% |
| MBPP+ | 72.2% | 72.2% |
| SWE-bench Verified | 31% | — |
| SWE-bench Verified (bash only) | 21.6% | — |
| GSO | 0% | — |
| CadEval | 26% | — |
Agentic & Tool Use GPT-4o mini leads
GPT-4o: 21.0 (#141), GPT-4o mini: 27.5 (#101)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| BALROG | 32.3% | 17.4% |
| GDPval | 9.9% | — |
| TheAgentCompany | 8.6% | — |
| Cybench | 12.5% | — |
| LMArena Search | 1006 | — |
| METR Time Horizons | 40.8% | — |
Reasoning Too close to call
GPT-4o: 9.4 (#343), GPT-4o mini: 8.7 (#347)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| ARC-AGI-2 | 0% | 0% |
| SimpleBench | 17.8% | 10.7% |
| Chess Puzzles | 13% | 0% |
| LiveBench Reasoning | 55.8% | 32.8% |
| LMArena Hard Prompts | 1281 | 1267 |
| DTBench | 64.5% | 54.4% |
| LiveBench Data Analysis | 60.9% | 50% |
| LMCA | 16.6% | 10.4% |
| Epoch Capabilities Index | 128.97 | 126.56 |
| LiveBench | 55.3% | 41.3% |
| Kagi LLM Benchmark | — | 28.8% |
| ARC-AGI-1 | 4.5% | — |
| CritPt | 0% | — |
| EnigmaEval | 0.8% | — |
| Mystery Game Puzzles | — | 12% |
| ForecastBench | 57.7 | — |
| PIQA | — | 88.7% |
Math Too close to call
GPT-4o: 10.6 (#312), GPT-4o mini: 10.4 (#314)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 0.4% | 0.7% |
| OTIS Mock AIME 2024-2025 | 6.4% | 6.9% |
| Omni-MATH | 29.3% | 28% |
| LiveBench Math | 49.5% | 36.3% |
| LMArena Math | 1285 | 1267 |
| MATH Level 5 | 53.3% | 52.6% |
| FrontierMath (Feb 2025 set) | 0.3% | — |
| GSM8K | — | 91.3% |
Knowledge GPT-4o leads
GPT-4o: 28.8 (#242), GPT-4o mini: 17.7 (#284)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| GPQA Diamond | 49.2% | 37.7% |
| SimpleQA Verified | 26% | 8.3% |
| MMLU-Pro | 71.3% | 60.3% |
| Confabulations | 15.3% | 37.2% |
| GPQA (HELM) | 52% | 36.8% |
| LMArena Expert | 1250 | 1235 |
| MMLU | 88.1% | 81.8% |
| Humanity's Last Exam | 2.7% | — |
| Vectara Hallucination Rate | 9.6% | — |
| BoolQ | — | 88.7% |
Multimodal GPT-4o leads
GPT-4o: 34.5 (#91), GPT-4o mini: 25.9 (#122)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| LMArena Vision | 1137 | 1066 |
| Video-MME | 71.9% | 64.8% |
| GeoBench | 71% | 64% |
| VPCT | 40% | 34% |
| ScienceQA | 88.5% | — |
Multilingual GPT-4o leads
GPT-4o: 43.2 (#186), GPT-4o mini: 42.0 (#199)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| LMArena Non-English | 1283 | 1266 |
| LMArena Chinese | 1277 | 1265 |
| LMArena French | 1304 | 1297 |
| LMArena German | 1282 | 1272 |
| LMArena Japanese | 1257 | 1216 |
| LMArena Korean | 1234 | 1195 |
| LMArena Russian | 1286 | 1275 |
| LMArena Spanish | 1292 | 1276 |
Instruction Following GPT-4o leads
GPT-4o: 66.6 (#207), GPT-4o mini: 61.9 (#239)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| LiveBench Instruction Following | 68.6% | 56.8% |
| IFEval | 81.7% | 78.2% |
| LMArena Instruction Following | 1278 | 1258 |
Long Context Too close to call
GPT-4o: 39.4 (#179), GPT-4o mini: 39.1 (#186)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| LMArena Longer Query | 1289 | 1289 |
| Fiction.LiveBench | 66.7% | — |
Writing & Preference GPT-4o leads
GPT-4o: 52.6 (#166), GPT-4o mini: 39.5 (#248)
| Benchmark | GPT-4o | GPT-4o mini |
|---|---|---|
| LMArena Text | 1300 | 1286 |
| LMArena Creative Writing | 1292 | 1268 |
| Short-Story Creative Writing | 81.8% | 67.2% |
| WildBench | 82.8% | 79.1% |
| LMArena Multi-Turn | 1302 | 1285 |
| LiveBench Language | 47.6% | 28.6% |
| EQ-Bench Creative Writing | — | 873 |
Frequently asked questions
Is GPT-4o better than GPT-4o mini?
GPT-4o is the stronger model overall, scoring 28.6 to 25.5 on the Noometry Index. GPT-4o mini costs 17× less per token, which makes it the better buy when GPT-4o's lead doesn't matter for your workload.
Which is cheaper, GPT-4o or GPT-4o mini?
GPT-4o mini is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; GPT-4o lists at $2.50 and $10.
Is GPT-4o or GPT-4o mini better for coding?
GPT-4o scores higher on coding benchmarks: 24.8 versus 22.0 in the Noometry coding category.
Which has the bigger context window?
Both accept 128K tokens.
How many benchmarks do GPT-4o and GPT-4o mini share?
54 benchmarks have published results for both models. GPT-4o has 72 scored results on Noometry and GPT-4o mini has 60.