Model comparison
GPT-4o vs GPT-5
GPT-5 is the stronger model overall, scoring 50.9 to 28.6 on the Noometry Index.
Last verified . 55 shared benchmarks.
Summary
- They share 55 benchmarks with published results for both. GPT-4o scores higher in 0 categories and GPT-5 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5 leads 55.0 to 10.6.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 6.4% for GPT-4o and 91.4% for GPT-5.
- GPT-5 is cheaper at $1.25 / $10 per million input/output tokens, against $2.50 / $10 for GPT-4o.
- GPT-5 accepts more context: 400K tokens versus 128K.
Side by side
| GPT-4o | GPT-5 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 28.6 | 50.9 |
| Released | 2024-05-13 | 2025-08-07 |
| Weights | Proprietary | Proprietary |
| Context window | 128K | 400K |
| Max output | 16K | 128K |
| Input $ / M tokens | $2.50 | $1.25 |
| Output $ / M tokens | $10 | $10 |
| Results tracked | 72 | 69 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 leads
GPT-4o: 24.8 (#328), GPT-5: 50.3 (#47)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| SWE-bench Verified | 31% | 73.6% |
| SWE-bench Verified (bash only) | 21.6% | 65% |
| Aider Polyglot | 45.3% | 88% |
| GSO | 0% | 6.9% |
| WeirdML | 25.1% | 60.7% |
| LMArena Coding | 1297 | 1436 |
| LMArena WebDev | — | 1418 |
| SciCode | — | 42.9% |
| BigCodeBench Instruct | 51.1% | — |
| LiveBench Coding | 51.4% | — |
| BigCodeBench Complete | 61.1% | — |
| CadEval | 26% | — |
| ALE-Bench | — | 1,162 |
| AlgoTune | — | 1.67 |
| HumanEval+ | 87.2% | — |
| MBPP+ | 72.2% | — |
Agentic & Tool Use GPT-5 leads
GPT-4o: 21.0 (#141), GPT-5: 33.1 (#56)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| GDPval | 9.9% | 34.8% |
| BALROG | 32.3% | 32.8% |
| LMArena Search | 1006 | 1133 |
| METR Time Horizons | 40.8% | 69.6% |
| Terminal-Bench | — | 49.6% |
| Remote Labor Index | — | 1.7% |
| TheAgentCompany | 8.6% | — |
| Cybench | 12.5% | — |
| DeepResearch Bench | — | 49.6% |
Reasoning GPT-5 leads
GPT-4o: 9.4 (#343), GPT-5: 38.3 (#64)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| ARC-AGI-2 | 0% | 9.9% |
| SimpleBench | 17.8% | 56.7% |
| ARC-AGI-1 | 4.5% | 65.7% |
| CritPt | 0% | 12.6% |
| Chess Puzzles | 13% | 37% |
| EnigmaEval | 0.8% | 10.5% |
| LMArena Hard Prompts | 1281 | 1416 |
| DTBench | 64.5% | 90.7% |
| LMCA | 16.6% | 40% |
| Epoch Capabilities Index | 128.97 | 150 |
| ForecastBench | 57.7 | 61.4 |
| Kagi LLM Benchmark | — | 72.7% |
| EBR-Bench | — | 12.7% |
| LiveBench Reasoning | 55.8% | — |
| Mystery Game Puzzles | — | 23% |
| LiveBench Data Analysis | 60.9% | — |
| LiveBench | 55.3% | — |
Math GPT-5 leads
GPT-4o: 10.6 (#312), GPT-5: 55.0 (#44)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 0.4% | 55.4% |
| OTIS Mock AIME 2024-2025 | 6.4% | 91.4% |
| Omni-MATH | 29.3% | 64.7% |
| LMArena Math | 1285 | 1407 |
| MATH Level 5 | 53.3% | 98.1% |
| FrontierMath (Feb 2025 set) | 0.3% | 32.4% |
| FrontierMath Tier 4 | — | 22% |
| ProofBench | — | 18% |
| LiveBench Math | 49.5% | — |
| FrontierMath Tier 4 (v1) | — | 12.5% |
Knowledge GPT-5 leads
GPT-4o: 28.8 (#242), GPT-5: 56.6 (#43)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| GPQA Diamond | 49.2% | 86.2% |
| Humanity's Last Exam | 2.7% | 25.3% |
| SimpleQA Verified | 26% | 50.1% |
| MMLU-Pro | 71.3% | 86.3% |
| Confabulations | 15.3% | 10.3% |
| Vectara Hallucination Rate | 9.6% | 14.7% |
| GPQA (HELM) | 52% | 79.2% |
| LMArena Expert | 1250 | 1419 |
| MMLU | 88.1% | — |
Multimodal GPT-5 leads
GPT-4o: 34.5 (#91), GPT-5: 46.8 (#13)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| LMArena Vision | 1137 | 1232 |
| GeoBench | 71% | 81% |
| VPCT | 40% | 66% |
| Video-MME | 71.9% | — |
| ScienceQA | 88.5% | — |
Multilingual GPT-5 leads
GPT-4o: 43.2 (#186), GPT-5: 51.4 (#110)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| LMArena Non-English | 1283 | 1397 |
| LMArena Chinese | 1277 | 1422 |
| LMArena French | 1304 | 1410 |
| LMArena German | 1282 | 1416 |
| LMArena Japanese | 1257 | 1409 |
| LMArena Korean | 1234 | 1360 |
| LMArena Russian | 1286 | 1406 |
| LMArena Spanish | 1292 | 1399 |
Instruction Following GPT-5 leads
GPT-4o: 66.6 (#207), GPT-5: 73.8 (#113)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| IFEval | 81.7% | 87.5% |
| LMArena Instruction Following | 1278 | 1388 |
| LiveBench Instruction Following | 68.6% | — |
Long Context GPT-5 leads
GPT-4o: 39.4 (#179), GPT-5: 69.5 (#2)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| Fiction.LiveBench | 66.7% | 97.2% |
| LMArena Longer Query | 1289 | 1399 |
Writing & Preference GPT-5 leads
GPT-4o: 52.6 (#166), GPT-5: 63.4 (#65)
| Benchmark | GPT-4o | GPT-5 |
|---|---|---|
| LMArena Text | 1300 | 1406 |
| LMArena Creative Writing | 1292 | 1365 |
| Short-Story Creative Writing | 81.8% | 86% |
| WildBench | 82.8% | 85.7% |
| LMArena Multi-Turn | 1302 | 1426 |
| EQ-Bench Creative Writing | — | 1627 |
| LiveBench Language | 47.6% | — |
Frequently asked questions
Is GPT-4o better than GPT-5?
GPT-5 is the stronger model overall, scoring 50.9 to 28.6 on the Noometry Index.
Which is cheaper, GPT-4o or GPT-5?
GPT-5 is cheaper. It lists at $1.25 per million input tokens and $10 per million output tokens; GPT-4o lists at $2.50 and $10.
Is GPT-4o or GPT-5 better for coding?
GPT-5 scores higher on coding benchmarks: 50.3 versus 24.8 in the Noometry coding category.
Which has the bigger context window?
GPT-5 does, with 400K tokens against 128K.
How many benchmarks do GPT-4o and GPT-5 share?
55 benchmarks have published results for both models. GPT-4o has 72 scored results on Noometry and GPT-5 has 69.