Model comparison
GPT-4 vs GPT-5 Mini
GPT-5 Mini is the stronger model overall, scoring 41.7 to 28.9 on the Noometry Index.
Last verified . 28 shared benchmarks.
Summary
- They share 28 benchmarks with published results for both. GPT-4 scores higher in 0 categories and GPT-5 Mini in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5 Mini leads 46.3 to 10.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.1% for GPT-4 and 86.7% for GPT-5 Mini.
- GPT-5 Mini is cheaper at $0.25 / $2 per million input/output tokens, against $30 / $60 for GPT-4.
- GPT-5 Mini accepts more context: 400K tokens versus 8K.
Side by side
| GPT-4 | GPT-5 Mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 28.9 | 41.7 |
| Released | 2023-03-14 | 2025-08-07 |
| Weights | Proprietary | Proprietary |
| Context window | 8K | 400K |
| Max output | 8K | 128K |
| Input $ / M tokens | $30 | $0.25 |
| Output $ / M tokens | $60 | $2 |
| Results tracked | 38 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 Mini leads
GPT-4: 31.2 (#277), GPT-5 Mini: 40.0 (#134)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| WeirdML | 12.4% | 52.7% |
| LMArena Coding | 1254 | 1406 |
| SWE-bench Verified | — | 64.7% |
| SWE-bench Verified (bash only) | — | 59.8% |
| SWE-bench Multilingual | — | 39.7% |
| SciCode | — | 39.2% |
| BigCodeBench Instruct | 46% | — |
| BigCodeBench Complete | 57.2% | — |
| ALE-Bench | — | 799.77 |
| AlgoTune | — | 1.38 |
| HumanEval+ | 79.3% | — |
Agentic & Tool Use Not comparable
GPT-4: —, GPT-5 Mini: 31.0 (#64)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| Terminal-Bench | — | 34.8% |
| Berkeley Function Calling Leaderboard | — | 55.5% |
| METR Time Horizons | 36.1% | — |
| Vending-Bench 2 | — | -31.18 |
Reasoning GPT-5 Mini leads
GPT-4: 17.3 (#285), GPT-5 Mini: 24.1 (#153)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| Chess Puzzles | 4% | 30% |
| LMArena Hard Prompts | 1241 | 1380 |
| Mystery Game Puzzles | 12% | 10% |
| DTBench | 62.7% | 80.5% |
| LMCA | 17.1% | 34.2% |
| Epoch Capabilities Index | 125.88 | 145.51 |
| ForecastBench | 57.8 | 61 |
| ARC-AGI-2 | — | 4.4% |
| Kagi LLM Benchmark | — | 70.3% |
| ARC-AGI-1 | — | 54.3% |
| CritPt | — | 0% |
| EnigmaEval | — | 8.2% |
| BIG-Bench Hard | 75.1% | — |
| HellaSwag | 95.3% | — |
| WinoGrande | 87.5% | — |
Math GPT-5 Mini leads
GPT-4: 10.1 (#304), GPT-5 Mini: 46.3 (#59)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 86.7% |
| LMArena Math | 1269 | 1377 |
| MATH Level 5 | 23% | 97.8% |
| FrontierMath (Tiers 1-3) | — | 46.7% |
| FrontierMath Tier 4 | — | 12.2% |
| ProofBench | — | 9% |
| Omni-MATH | — | 72.2% |
| FrontierMath (Feb 2025 set) | — | 27.2% |
| FrontierMath Tier 4 (v1) | — | 6.3% |
| GSM8K | 92% | — |
Knowledge GPT-5 Mini leads
GPT-4: 18.4 (#277), GPT-5 Mini: 45.5 (#80)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| GPQA Diamond | 35.7% | 75% |
| LMArena Expert | 1211 | 1377 |
| Humanity's Last Exam | — | 19.4% |
| SimpleQA Verified | — | 21.6% |
| MMLU-Pro | — | 83.5% |
| Confabulations | — | 13.3% |
| Vectara Hallucination Rate | — | 12.9% |
| GPQA (HELM) | — | 75.6% |
| MMLU | 86.4% | — |
| TriviaQA | 84.8% | — |
Multimodal Not comparable
GPT-4: —, GPT-5 Mini: 35.0 (#80)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| LMArena Vision | — | 1202 |
| VPCT | — | 40.2% |
Multilingual GPT-5 Mini leads
GPT-4: 40.8 (#210), GPT-5 Mini: 49.1 (#131)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| LMArena Non-English | 1246 | 1363 |
| LMArena Chinese | 1242 | 1387 |
| LMArena French | 1282 | 1385 |
| LMArena German | 1251 | 1366 |
| LMArena Japanese | 1209 | 1342 |
| LMArena Korean | 1184 | 1306 |
| LMArena Russian | 1251 | 1362 |
| LMArena Spanish | 1261 | 1355 |
Instruction Following GPT-5 Mini leads
GPT-4: 65.5 (#217), GPT-5 Mini: 76.4 (#43)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| LMArena Instruction Following | 1241 | 1357 |
| IFEval | — | 92.7% |
Long Context GPT-5 Mini leads
GPT-4: 37.7 (#207), GPT-5 Mini: 41.9 (#127)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| LMArena Longer Query | 1244 | 1355 |
| Fiction.LiveBench | — | 69.4% |
Writing & Preference GPT-5 Mini leads
GPT-4: 35.0 (#263), GPT-5 Mini: 55.4 (#143)
| Benchmark | GPT-4 | GPT-5 Mini |
|---|---|---|
| LMArena Text | 1263 | 1373 |
| LMArena Creative Writing | 1244 | 1325 |
| EQ-Bench Creative Writing | 752 | 1313 |
| LMArena Multi-Turn | 1257 | 1363 |
| Short-Story Creative Writing | — | 83.1% |
| WildBench | — | 85.5% |
Frequently asked questions
Is GPT-4 better than GPT-5 Mini?
GPT-5 Mini is the stronger model overall, scoring 41.7 to 28.9 on the Noometry Index.
Which is cheaper, GPT-4 or GPT-5 Mini?
GPT-5 Mini is cheaper. It lists at $0.25 per million input tokens and $2 per million output tokens; GPT-4 lists at $30 and $60.
Is GPT-4 or GPT-5 Mini better for coding?
GPT-5 Mini scores higher on coding benchmarks: 40.0 versus 31.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5 Mini does, with 400K tokens against 8K.
How many benchmarks do GPT-4 and GPT-5 Mini share?
28 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and GPT-5 Mini has 60.