Model comparison
GPT-5 vs GPT-5.4
GPT-5.4 is the stronger model overall, scoring 59.4 to 50.9 on the Noometry Index. GPT-5 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 . 52 shared benchmarks.
Summary
- They share 52 benchmarks with published results for both. GPT-5 scores higher in 2 categories and GPT-5.4 in 8 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.8 to 38.3.
- The biggest single-benchmark swing is ARC-AGI-2: 9.9% for GPT-5 and 74% for GPT-5.4.
- GPT-5 is cheaper at $1.25 / $10 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 400K.
Side by side
| GPT-5 | GPT-5.4 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 50.9 | 59.4 |
| Released | 2025-08-07 | 2026-03-05 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $1.25 | $2.50 |
| Output $ / M tokens | $10 | $15 |
| Results tracked | 69 | 68 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
GPT-5: 50.3 (#47), GPT-5.4: 52.6 (#33)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| SWE-bench Verified | 73.6% | 76.9% |
| LMArena WebDev | 1418 | 1465 |
| SciCode | 42.9% | 56.6% |
| GSO | 6.9% | 31.4% |
| WeirdML | 60.7% | 77.7% |
| LMArena Coding | 1436 | 1497 |
| ALE-Bench | 1,162 | 1,607 |
| AlgoTune | 1.67 | 1.85 |
| DeepSWE | — | 51.8% |
| SWE-bench Verified (bash only) | 65% | — |
| Aider Polyglot | 88% | — |
| MirrorCode | — | 15.6% |
Agentic & Tool Use GPT-5.4 leads
GPT-5: 33.1 (#56), GPT-5.4: 46.5 (#13)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| Terminal-Bench | 49.6% | 81.8% |
| DeepResearch Bench | 49.6% | 35.1% |
| LMArena Search | 1133 | 1197 |
| METR Time Horizons | 69.6% | 74.3% |
| APEX-Agents | — | 52.4% |
| GDPval | 34.8% | — |
| Remote Labor Index | 1.7% | — |
| τ²-bench Banking | — | 39.4% |
| PostTrainBench | — | 19% |
| BALROG | 32.8% | — |
| GBAEval | — | 45.1% |
| Vending-Bench 2 | — | 6,144 |
Reasoning GPT-5.4 leads
GPT-5: 38.3 (#64), GPT-5.4: 61.8 (#19)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| ARC-AGI-2 | 9.9% | 74% |
| Kagi LLM Benchmark | 72.7% | 63.8% |
| ARC-AGI-1 | 65.7% | 93.7% |
| CritPt | 12.6% | 23.4% |
| Chess Puzzles | 37% | 44% |
| EnigmaEval | 10.5% | 16% |
| EBR-Bench | 12.7% | 25.4% |
| LMArena Hard Prompts | 1416 | 1485 |
| Mystery Game Puzzles | 23% | 37% |
| DTBench | 90.7% | 94.4% |
| LMCA | 40% | 52% |
| Epoch Capabilities Index | 150 | 156.81 |
| ForecastBench | 61.4 | 59.5 |
| SimpleBench | 56.7% | — |
| NYT Connections (extended) | — | 91.3% |
| Thematic Generalization | — | 80% |
Math GPT-5.4 leads
GPT-5: 55.0 (#44), GPT-5.4: 73.5 (#19)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 55.4% | 78.6% |
| FrontierMath Tier 4 | 22% | 49% |
| OTIS Mock AIME 2024-2025 | 91.4% | 97.8% |
| ProofBench | 18% | 56% |
| LMArena Math | 1407 | 1488 |
| FrontierMath (Feb 2025 set) | 32.4% | 47.6% |
| FrontierMath Tier 4 (v1) | 12.5% | 27.1% |
| MathArena Final-Answer Competitions | — | 83.1% |
| Omni-MATH | 64.7% | — |
| MATH Level 5 | 98.1% | — |
Knowledge GPT-5.4 leads
GPT-5: 56.6 (#43), GPT-5.4: 65.3 (#14)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 86.2% | 93.3% |
| Humanity's Last Exam | 25.3% | 36.2% |
| SimpleQA Verified | 50.1% | 45.1% |
| Vectara Hallucination Rate | 14.7% | 7% |
| LMArena Expert | 1419 | 1507 |
| MMLU-Pro | 86.3% | — |
| Confabulations | 10.3% | — |
| GPQA (HELM) | 79.2% | — |
Multimodal GPT-5 leads
GPT-5: 46.8 (#13), GPT-5.4: 43.7 (#20)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| LMArena Vision | 1232 | 1303 |
| GeoBench | 81% | — |
| VPCT | 66% | — |
| Blueprint-Bench 2 | — | 27.1% |
| Furniture Assembly | — | 37.5% |
| LMArena Document | — | 1471 |
Multilingual GPT-5.4 leads
GPT-5: 51.4 (#110), GPT-5.4: 56.2 (#23)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1397 | 1465 |
| LMArena Chinese | 1422 | 1519 |
| LMArena French | 1410 | 1493 |
| LMArena German | 1416 | 1472 |
| LMArena Japanese | 1409 | 1485 |
| LMArena Korean | 1360 | 1448 |
| LMArena Russian | 1406 | 1480 |
| LMArena Spanish | 1399 | 1454 |
Instruction Following GPT-5.4 leads
GPT-5: 73.8 (#113), GPT-5.4: 77.1 (#27)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1388 | 1469 |
| IFEval | 87.5% | — |
Long Context GPT-5 leads
GPT-5: 69.5 (#2), GPT-5.4: 50.3 (#8)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1399 | 1473 |
| Fiction.LiveBench | 97.2% | — |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |
Writing & Preference GPT-5.4 leads
GPT-5: 63.4 (#65), GPT-5.4: 71.9 (#17)
| Benchmark | GPT-5 | GPT-5.4 |
|---|---|---|
| LMArena Text | 1406 | 1469 |
| LMArena Creative Writing | 1365 | 1439 |
| EQ-Bench Creative Writing | 1627 | 1840 |
| LMArena Multi-Turn | 1426 | 1482 |
| Short-Story Creative Writing | 86% | — |
| WildBench | 85.7% | — |
| EQ-Bench 4 | — | 1272 |
Frequently asked questions
Is GPT-5 better than GPT-5.4?
GPT-5.4 is the stronger model overall, scoring 59.4 to 50.9 on the Noometry Index. GPT-5 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 or GPT-5.4?
GPT-5 is cheaper. It lists at $1.25 per million input tokens and $10 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is GPT-5 or GPT-5.4 better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.6 versus 50.3 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 400K.
How many benchmarks do GPT-5 and GPT-5.4 share?
52 benchmarks have published results for both models. GPT-5 has 69 scored results on Noometry and GPT-5.4 has 68.