Model comparison
GPT-4 vs GPT-4o mini
GPT-4 is the stronger model overall, scoring 29.1 to 25.5 on the Noometry Index. GPT-4o mini costs 143× less per token, which makes it the better buy when GPT-4's lead doesn't matter for your workload.
Last verified . 32 shared benchmarks.
Summary
- They share 32 benchmarks with published results for both. GPT-4 scores higher in 5 categories and GPT-4o mini in 3 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in coding, where GPT-4 leads 31.6 to 22.0.
- The biggest single-benchmark swing is MATH Level 5: 23% for GPT-4 and 52.6% for GPT-4o mini.
- GPT-4o mini is cheaper at $0.15 / $0.60 per million input/output tokens, against $30 / $60 for GPT-4.
- GPT-4o mini accepts more context: 128K tokens versus 8K.
Side by side
| GPT-4 | GPT-4o mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 29.1 | 25.5 |
| Released | 2023-03-14 | 2024-07-18 |
| Weights | Proprietary | Proprietary |
| Context window | 8K | 128K |
| Max output | 8K | 16K |
| Input $ / M tokens | $30 | $0.15 |
| Output $ / M tokens | $60 | $0.60 |
| Results tracked | 38 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-4 leads
GPT-4: 31.6 (#283), GPT-4o mini: 22.0 (#335)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| WeirdML | 12.4% | 11.8% |
| BigCodeBench Instruct | 46% | 46.1% |
| LMArena Coding | 1254 | 1290 |
| BigCodeBench Complete | 57.2% | 57.4% |
| HumanEval+ | 79.3% | 83.5% |
| Aider Polyglot | — | 3.6% |
| LiveBench Coding | — | 43.1% |
| MBPP+ | — | 72.2% |
Agentic & Tool Use Not comparable
GPT-4: —, GPT-4o mini: 27.5 (#101)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| BALROG | — | 17.4% |
| METR Time Horizons | 36.1% | — |
Reasoning GPT-4 leads
GPT-4: 17.8 (#289), GPT-4o mini: 8.7 (#347)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| Chess Puzzles | 4% | 0% |
| LMArena Hard Prompts | 1241 | 1267 |
| Mystery Game Puzzles | 12% | 12% |
| DTBench | 62.7% | 54.4% |
| LMCA | 17.1% | 10.4% |
| Epoch Capabilities Index | 125.89 | 126.56 |
| ARC-AGI-2 | — | 0% |
| SimpleBench | — | 10.7% |
| Kagi LLM Benchmark | — | 28.8% |
| LiveBench Reasoning | — | 32.8% |
| LiveBench Data Analysis | — | 50% |
| BIG-Bench Hard | 75.1% | — |
| ForecastBench | 57.8 | — |
| HellaSwag | 95.3% | — |
| LiveBench | — | 41.3% |
| PIQA | — | 88.7% |
| WinoGrande | 87.5% | — |
Math Too close to call
GPT-4: 10.8 (#309), GPT-4o mini: 10.4 (#314)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 6.9% |
| LMArena Math | 1269 | 1267 |
| MATH Level 5 | 23% | 52.6% |
| GSM8K | 92% | 91.3% |
| FrontierMath (Tiers 1-3) | — | 0.7% |
| Omni-MATH | — | 28% |
| LiveBench Math | — | 36.3% |
Knowledge Too close to call
GPT-4: 18.4 (#282), GPT-4o mini: 17.7 (#284)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| GPQA Diamond | 35.7% | 37.7% |
| LMArena Expert | 1211 | 1235 |
| MMLU | 86.4% | 81.8% |
| SimpleQA Verified | — | 8.3% |
| MMLU-Pro | — | 60.3% |
| Confabulations | — | 37.2% |
| GPQA (HELM) | — | 36.8% |
| BoolQ | — | 88.7% |
| TriviaQA | 84.8% | — |
Multimodal Not comparable
GPT-4: —, GPT-4o mini: 25.9 (#122)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| LMArena Vision | — | 1066 |
| Video-MME | — | 64.8% |
| GeoBench | — | 64% |
| VPCT | — | 34% |
Multilingual GPT-4o mini leads
GPT-4: 40.6 (#215), GPT-4o mini: 42.0 (#199)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| LMArena Non-English | 1246 | 1266 |
| LMArena Chinese | 1242 | 1265 |
| LMArena French | 1283 | 1297 |
| LMArena German | 1251 | 1272 |
| LMArena Japanese | 1209 | 1216 |
| LMArena Korean | 1184 | 1195 |
| LMArena Russian | 1251 | 1275 |
| LMArena Spanish | 1261 | 1276 |
Instruction Following GPT-4 leads
GPT-4: 65.3 (#222), GPT-4o mini: 61.9 (#239)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| LMArena Instruction Following | 1241 | 1258 |
| LiveBench Instruction Following | — | 56.8% |
| IFEval | — | 78.2% |
Long Context GPT-4o mini leads
GPT-4: 37.7 (#212), GPT-4o mini: 39.1 (#186)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| LMArena Longer Query | 1244 | 1289 |
Writing & Preference GPT-4o mini leads
GPT-4: 34.9 (#268), GPT-4o mini: 39.5 (#248)
| Benchmark | GPT-4 | GPT-4o mini |
|---|---|---|
| LMArena Text | 1263 | 1286 |
| LMArena Creative Writing | 1244 | 1268 |
| EQ-Bench Creative Writing | 752 | 873 |
| LMArena Multi-Turn | 1257 | 1285 |
| Short-Story Creative Writing | — | 67.2% |
| WildBench | — | 79.1% |
| LiveBench Language | — | 28.6% |
Frequently asked questions
Is GPT-4 better than GPT-4o mini?
GPT-4 is the stronger model overall, scoring 29.1 to 25.5 on the Noometry Index. GPT-4o mini costs 143× less per token, which makes it the better buy when GPT-4's lead doesn't matter for your workload.
Which is cheaper, GPT-4 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-4 lists at $30 and $60.
Is GPT-4 or GPT-4o mini better for coding?
GPT-4 scores higher on coding benchmarks: 31.6 versus 22.0 in the Noometry coding category.
Which has the bigger context window?
GPT-4o mini does, with 128K tokens against 8K.
How many benchmarks do GPT-4 and GPT-4o mini share?
32 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and GPT-4o mini has 60.