Model comparison
GPT-5.4 vs o3
GPT-5.4 is the stronger model overall, scoring 59.4 to 47.5 on the Noometry Index. o3 costs 1.6× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 45 shared benchmarks.
Summary
- They share 45 benchmarks with published results for both. GPT-5.4 scores higher in 9 categories and o3 in 1 category; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.8 to 32.0.
- The biggest single-benchmark swing is ARC-AGI-2: 74% for GPT-5.4 and 6.5% for o3.
- o3 is cheaper at $2 / $8 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 200K.
Side by side
| GPT-5.4 | o3 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 59.4 | 47.5 |
| Released | 2026-03-05 | 2025-04-16 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 200K |
| Max output | 128K | 100K |
| Input $ / M tokens | $2.50 | $2 |
| Output $ / M tokens | $15 | $8 |
| Results tracked | 68 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
GPT-5.4: 52.6 (#33), o3: 46.8 (#64)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| SWE-bench Verified | 76.9% | 62.3% |
| GSO | 31.4% | 8.8% |
| WeirdML | 77.7% | 52.4% |
| LMArena Coding | 1497 | 1408 |
| ALE-Bench | 1,607 | 933.55 |
| DeepSWE | 51.8% | — |
| SWE-bench Verified (bash only) | — | 58.4% |
| Aider Polyglot | — | 81.3% |
| LMArena WebDev | 1465 | — |
| SciCode | 56.6% | — |
| MirrorCode | 15.6% | — |
| CadEval | — | 74% |
| AlgoTune | 1.85 | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#13), o3: 34.5 (#44)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| DeepResearch Bench | 35.1% | 45.2% |
| LMArena Search | 1197 | 1144 |
| METR Time Horizons | 74.3% | 65.4% |
| Terminal-Bench | 81.8% | — |
| APEX-Agents | 52.4% | — |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 30.8% |
| τ²-bench Banking | 39.4% | — |
| OSWorld | — | 23% |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| Vending-Bench 2 | 6,144 | — |
Reasoning GPT-5.4 leads
GPT-5.4: 61.8 (#19), o3: 32.0 (#78)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| ARC-AGI-2 | 74% | 6.5% |
| Kagi LLM Benchmark | 63.8% | 67.6% |
| ARC-AGI-1 | 93.7% | 60.8% |
| CritPt | 23.4% | 1.4% |
| Chess Puzzles | 44% | 38% |
| EnigmaEval | 16% | 13.1% |
| LMArena Hard Prompts | 1485 | 1402 |
| Mystery Game Puzzles | 37% | 29% |
| DTBench | 94.4% | 84.8% |
| LMCA | 52% | 39.7% |
| Epoch Capabilities Index | 156.81 | 146.86 |
| ForecastBench | 59.5 | 62.5 |
| SimpleBench | — | 53.1% |
| NYT Connections (extended) | 91.3% | — |
| Thematic Generalization | 80% | — |
| EBR-Bench | 25.4% | — |
Math GPT-5.4 leads
GPT-5.4: 73.5 (#19), o3: 50.2 (#58)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 78.6% | 33.3% |
| OTIS Mock AIME 2024-2025 | 97.8% | 84.4% |
| LMArena Math | 1488 | 1426 |
| FrontierMath (Feb 2025 set) | 47.6% | 18.7% |
| FrontierMath Tier 4 (v1) | 27.1% | 2.1% |
| FrontierMath Tier 4 | 49% | — |
| MathArena Final-Answer Competitions | 83.1% | — |
| ProofBench | 56% | — |
| Omni-MATH | — | 71.4% |
| MATH Level 5 | — | 97.8% |
Knowledge GPT-5.4 leads
GPT-5.4: 65.3 (#14), o3: 54.6 (#52)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| GPQA Diamond | 93.3% | 81.8% |
| Humanity's Last Exam | 36.2% | 20.3% |
| SimpleQA Verified | 45.1% | 49.4% |
| LMArena Expert | 1507 | 1402 |
| MMLU-Pro | — | 85.9% |
| Confabulations | — | 14.4% |
| Vectara Hallucination Rate | 7% | — |
| GPQA (HELM) | — | 75.3% |
Multimodal GPT-5.4 leads
GPT-5.4: 43.7 (#20), o3: 41.4 (#36)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| LMArena Vision | 1303 | 1214 |
| GeoBench | — | 74% |
| VPCT | — | 52% |
| Blueprint-Bench 2 | 27.1% | — |
| Furniture Assembly | 37.5% | — |
| LMArena Document | 1471 | — |
Multilingual GPT-5.4 leads
GPT-5.4: 56.2 (#23), o3: 51.7 (#105)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| LMArena Non-English | 1465 | 1401 |
| LMArena Chinese | 1519 | 1437 |
| LMArena French | 1493 | 1430 |
| LMArena German | 1472 | 1420 |
| LMArena Japanese | 1485 | 1403 |
| LMArena Korean | 1448 | 1370 |
| LMArena Russian | 1480 | 1406 |
| LMArena Spanish | 1454 | 1395 |
Instruction Following GPT-5.4 leads
GPT-5.4: 77.1 (#27), o3: 72.8 (#127)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| LMArena Instruction Following | 1469 | 1368 |
| IFEval | — | 86.9% |
Long Context o3 leads
GPT-5.4: 50.3 (#8), o3: 53.3 (#6)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| CL-bench | 27.9% | 17.8% |
| LMArena Longer Query | 1473 | 1372 |
| Fiction.LiveBench | — | 88.9% |
| CL-bench Life | 21.7% | — |
Writing & Preference GPT-5.4 leads
GPT-5.4: 71.9 (#17), o3: 63.5 (#64)
| Benchmark | GPT-5.4 | o3 |
|---|---|---|
| LMArena Text | 1469 | 1410 |
| LMArena Creative Writing | 1439 | 1359 |
| EQ-Bench Creative Writing | 1840 | 1676 |
| LMArena Multi-Turn | 1482 | 1405 |
| Short-Story Creative Writing | — | 83.9% |
| WildBench | — | 86.1% |
| EQ-Bench 4 | 1272 | — |
Frequently asked questions
Is GPT-5.4 better than o3?
GPT-5.4 is the stronger model overall, scoring 59.4 to 47.5 on the Noometry Index. o3 costs 1.6× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 or o3?
o3 is cheaper. It lists at $2 per million input tokens and $8 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is GPT-5.4 or o3 better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.6 versus 46.8 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 200K.
How many benchmarks do GPT-5.4 and o3 share?
45 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and o3 has 63.