Model comparison
GPT-4 Turbo vs o3
o3 is the stronger model overall, scoring 47.5 to 30.5 on the Noometry Index.
Last verified . 31 shared benchmarks.
Summary
- They share 31 benchmarks with published results for both. GPT-4 Turbo scores higher in 0 categories and o3 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where o3 leads 50.2 to 9.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 6.7% for GPT-4 Turbo and 84.4% for o3.
- o3 is cheaper at $2 / $8 per million input/output tokens, against $10 / $30 for GPT-4 Turbo.
- o3 accepts more context: 200K tokens versus 128K.
Side by side
| GPT-4 Turbo | o3 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 30.5 | 47.5 |
| Released | 2023-11-06 | 2025-04-16 |
| Weights | Proprietary | Proprietary |
| Context window | 128K | 200K |
| Max output | 4K | 100K |
| Input $ / M tokens | $10 | $2 |
| Output $ / M tokens | $30 | $8 |
| Results tracked | 36 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o3 leads
GPT-4 Turbo: 33.8 (#249), o3: 46.8 (#64)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| WeirdML | 18% | 52.4% |
| LMArena Coding | 1268 | 1408 |
| SWE-bench Verified | — | 62.3% |
| SWE-bench Verified (bash only) | — | 58.4% |
| Aider Polyglot | — | 81.3% |
| GSO | — | 8.8% |
| BigCodeBench Instruct | 48.2% | — |
| BigCodeBench Complete | 58.2% | — |
| CadEval | — | 74% |
| ALE-Bench | — | 933.55 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73.3% | — |
Agentic & Tool Use Not comparable
GPT-4 Turbo: —, o3: 34.5 (#44)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| METR Time Horizons | 36.7% | 65.4% |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 30.8% |
| DeepResearch Bench | — | 45.2% |
| OSWorld | — | 23% |
| LMArena Search | — | 1144 |
Reasoning o3 leads
GPT-4 Turbo: 15.3 (#317), o3: 32.0 (#78)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| SimpleBench | 25.1% | 53.1% |
| Chess Puzzles | 6% | 38% |
| LMArena Hard Prompts | 1251 | 1402 |
| DTBench | 61.6% | 84.8% |
| LMCA | 9.8% | 39.7% |
| Epoch Capabilities Index | 127.25 | 146.86 |
| ForecastBench | 59.4 | 62.5 |
| ARC-AGI-2 | — | 6.5% |
| Kagi LLM Benchmark | — | 67.6% |
| ARC-AGI-1 | — | 60.8% |
| CritPt | — | 1.4% |
| EnigmaEval | — | 13.1% |
| Mystery Game Puzzles | — | 29% |
Math o3 leads
GPT-4 Turbo: 9.0 (#322), o3: 50.2 (#58)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 0.7% | 33.3% |
| OTIS Mock AIME 2024-2025 | 6.7% | 84.4% |
| LMArena Math | 1272 | 1426 |
| MATH Level 5 | 46.7% | 97.8% |
| Omni-MATH | — | 71.4% |
| FrontierMath (Feb 2025 set) | — | 18.7% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge o3 leads
GPT-4 Turbo: 24.3 (#268), o3: 54.6 (#52)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| GPQA Diamond | 46.6% | 81.8% |
| Confabulations | 28.4% | 14.4% |
| LMArena Expert | 1223 | 1402 |
| Humanity's Last Exam | — | 20.3% |
| SimpleQA Verified | — | 49.4% |
| MMLU-Pro | — | 85.9% |
| GPQA (HELM) | — | 75.3% |
| MMLU | 81.3% | — |
Multimodal o3 leads
GPT-4 Turbo: 30.6 (#110), o3: 41.4 (#36)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| LMArena Vision | 1090 | 1214 |
| GeoBench | — | 74% |
| VPCT | — | 52% |
Multilingual o3 leads
GPT-4 Turbo: 40.5 (#216), o3: 51.7 (#105)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| LMArena Non-English | 1245 | 1401 |
| LMArena Chinese | 1242 | 1437 |
| LMArena French | 1276 | 1430 |
| LMArena German | 1259 | 1420 |
| LMArena Japanese | 1194 | 1403 |
| LMArena Korean | 1187 | 1370 |
| LMArena Russian | 1259 | 1406 |
| LMArena Spanish | 1260 | 1395 |
Instruction Following o3 leads
GPT-4 Turbo: 65.8 (#216), o3: 72.8 (#127)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| LMArena Instruction Following | 1249 | 1368 |
| IFEval | — | 86.9% |
Long Context o3 leads
GPT-4 Turbo: 38.0 (#206), o3: 53.3 (#6)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| LMArena Longer Query | 1254 | 1372 |
| Fiction.LiveBench | — | 88.9% |
| CL-bench | — | 17.8% |
Writing & Preference o3 leads
GPT-4 Turbo: 47.7 (#206), o3: 63.5 (#64)
| Benchmark | GPT-4 Turbo | o3 |
|---|---|---|
| LMArena Text | 1272 | 1410 |
| LMArena Creative Writing | 1269 | 1359 |
| LMArena Multi-Turn | 1267 | 1405 |
| Short-Story Creative Writing | — | 83.9% |
| EQ-Bench Creative Writing | — | 1676 |
| WildBench | — | 86.1% |
Frequently asked questions
Is GPT-4 Turbo better than o3?
o3 is the stronger model overall, scoring 47.5 to 30.5 on the Noometry Index.
Which is cheaper, GPT-4 Turbo or o3?
o3 is cheaper. It lists at $2 per million input tokens and $8 per million output tokens; GPT-4 Turbo lists at $10 and $30.
Is GPT-4 Turbo or o3 better for coding?
o3 scores higher on coding benchmarks: 46.8 versus 33.8 in the Noometry coding category.
Which has the bigger context window?
o3 does, with 200K tokens against 128K.
How many benchmarks do GPT-4 Turbo and o3 share?
31 benchmarks have published results for both models. GPT-4 Turbo has 36 scored results on Noometry and o3 has 63.