Model comparison
GPT-5 vs o3
GPT-5 is the stronger model overall, scoring 50.9 to 47.5 on the Noometry Index.
Last verified . 59 shared benchmarks.
Summary
- They share 59 benchmarks with published results for both. GPT-5 scores higher in 7 categories and o3 in 3 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-5 leads 69.5 to 53.3.
- The biggest single-benchmark swing is FrontierMath (Tiers 1-3): 55.4% for GPT-5 and 33.3% for o3.
- Both cost about the same: $1.25 input and $10 output per million tokens.
- GPT-5 accepts more context: 400K tokens versus 200K.
Side by side
| GPT-5 | o3 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 50.9 | 47.5 |
| Released | 2025-08-07 | 2025-04-16 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 200K |
| Max output | 128K | 100K |
| Input $ / M tokens | $1.25 | $2 |
| Output $ / M tokens | $10 | $8 |
| Results tracked | 69 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 leads
GPT-5: 50.3 (#47), o3: 46.8 (#64)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| SWE-bench Verified | 73.6% | 62.3% |
| SWE-bench Verified (bash only) | 65% | 58.4% |
| Aider Polyglot | 88% | 81.3% |
| GSO | 6.9% | 8.8% |
| WeirdML | 60.7% | 52.4% |
| LMArena Coding | 1436 | 1408 |
| ALE-Bench | 1,162 | 933.55 |
| LMArena WebDev | 1418 | — |
| SciCode | 42.9% | — |
| CadEval | — | 74% |
| AlgoTune | 1.67 | — |
Agentic & Tool Use o3 leads
GPT-5: 33.1 (#56), o3: 34.5 (#44)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| GDPval | 34.8% | 30.8% |
| DeepResearch Bench | 49.6% | 45.2% |
| LMArena Search | 1133 | 1144 |
| METR Time Horizons | 69.6% | 65.4% |
| Terminal-Bench | 49.6% | — |
| Berkeley Function Calling Leaderboard | — | 63% |
| Remote Labor Index | 1.7% | — |
| OSWorld | — | 23% |
| BALROG | 32.8% | — |
Reasoning GPT-5 leads
GPT-5: 38.3 (#64), o3: 32.0 (#78)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| ARC-AGI-2 | 9.9% | 6.5% |
| SimpleBench | 56.7% | 53.1% |
| Kagi LLM Benchmark | 72.7% | 67.6% |
| ARC-AGI-1 | 65.7% | 60.8% |
| CritPt | 12.6% | 1.4% |
| Chess Puzzles | 37% | 38% |
| EnigmaEval | 10.5% | 13.1% |
| LMArena Hard Prompts | 1416 | 1402 |
| Mystery Game Puzzles | 23% | 29% |
| DTBench | 90.7% | 84.8% |
| LMCA | 40% | 39.7% |
| Epoch Capabilities Index | 150 | 146.86 |
| ForecastBench | 61.4 | 62.5 |
| EBR-Bench | 12.7% | — |
Math GPT-5 leads
GPT-5: 55.0 (#44), o3: 50.2 (#58)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 55.4% | 33.3% |
| OTIS Mock AIME 2024-2025 | 91.4% | 84.4% |
| Omni-MATH | 64.7% | 71.4% |
| LMArena Math | 1407 | 1426 |
| MATH Level 5 | 98.1% | 97.8% |
| FrontierMath (Feb 2025 set) | 32.4% | 18.7% |
| FrontierMath Tier 4 (v1) | 12.5% | 2.1% |
| FrontierMath Tier 4 | 22% | — |
| ProofBench | 18% | — |
Knowledge GPT-5 leads
GPT-5: 56.6 (#43), o3: 54.6 (#52)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| GPQA Diamond | 86.2% | 81.8% |
| Humanity's Last Exam | 25.3% | 20.3% |
| SimpleQA Verified | 50.1% | 49.4% |
| MMLU-Pro | 86.3% | 85.9% |
| Confabulations | 10.3% | 14.4% |
| GPQA (HELM) | 79.2% | 75.3% |
| LMArena Expert | 1419 | 1402 |
| Vectara Hallucination Rate | 14.7% | — |
Multimodal GPT-5 leads
GPT-5: 46.8 (#13), o3: 41.4 (#36)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| LMArena Vision | 1232 | 1214 |
| GeoBench | 81% | 74% |
| VPCT | 66% | 52% |
Multilingual Too close to call
GPT-5: 51.4 (#110), o3: 51.7 (#105)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| LMArena Non-English | 1397 | 1401 |
| LMArena Chinese | 1422 | 1437 |
| LMArena French | 1410 | 1430 |
| LMArena German | 1416 | 1420 |
| LMArena Japanese | 1409 | 1403 |
| LMArena Korean | 1360 | 1370 |
| LMArena Russian | 1406 | 1406 |
| LMArena Spanish | 1399 | 1395 |
Instruction Following GPT-5 leads
GPT-5: 73.8 (#113), o3: 72.8 (#127)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| IFEval | 87.5% | 86.9% |
| LMArena Instruction Following | 1388 | 1368 |
Long Context GPT-5 leads
GPT-5: 69.5 (#2), o3: 53.3 (#6)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| Fiction.LiveBench | 97.2% | 88.9% |
| LMArena Longer Query | 1399 | 1372 |
| CL-bench | — | 17.8% |
Writing & Preference Too close to call
GPT-5: 63.4 (#65), o3: 63.5 (#64)
| Benchmark | GPT-5 | o3 |
|---|---|---|
| LMArena Text | 1406 | 1410 |
| LMArena Creative Writing | 1365 | 1359 |
| Short-Story Creative Writing | 86% | 83.9% |
| EQ-Bench Creative Writing | 1627 | 1676 |
| WildBench | 85.7% | 86.1% |
| LMArena Multi-Turn | 1426 | 1405 |
Frequently asked questions
Is GPT-5 better than o3?
GPT-5 is the stronger model overall, scoring 50.9 to 47.5 on the Noometry Index.
Which is cheaper, GPT-5 or o3?
GPT-5 is cheaper. It lists at $1.25 per million input tokens and $10 per million output tokens; o3 lists at $2 and $8.
Is GPT-5 or o3 better for coding?
GPT-5 scores higher on coding benchmarks: 50.3 versus 46.8 in the Noometry coding category.
Which has the bigger context window?
GPT-5 does, with 400K tokens against 200K.
How many benchmarks do GPT-5 and o3 share?
59 benchmarks have published results for both models. GPT-5 has 69 scored results on Noometry and o3 has 63.