Model comparison
GPT-4.1 vs GPT-5.4
GPT-5.4 is the stronger model overall, scoring 59.5 to 35.7 on the Noometry Index. GPT-4.1 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 . 39 shared benchmarks.
Summary
- They share 39 benchmarks with published results for both. GPT-4.1 scores higher in 0 categories and GPT-5.4 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 leads 73.3 to 21.4.
- The biggest single-benchmark swing is ARC-AGI-1: 5.5% for GPT-4.1 and 93.7% for GPT-5.4.
- GPT-4.1 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 1.05M.
Side by side
| GPT-4.1 | GPT-5.4 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 35.7 | 59.5 |
| Released | 2025-04-14 | 2026-03-05 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 33K | 128K |
| Input $ / M tokens | $2 | $2.50 |
| Output $ / M tokens | $8 | $15 |
| Results tracked | 52 | 68 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
GPT-4.1: 34.2 (#228), GPT-5.4: 52.8 (#28)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| SWE-bench Verified | 48.5% | 76.9% |
| WeirdML | 39% | 77.7% |
| LMArena Coding | 1391 | 1497 |
| ALE-Bench | 558.1 | 1,607 |
| DeepSWE | — | 51.8% |
| SWE-bench Verified (bash only) | 39.6% | — |
| Aider Polyglot | 52.4% | — |
| LMArena WebDev | — | 1465 |
| SciCode | — | 56.6% |
| GSO | — | 31.4% |
| MirrorCode | — | 15.6% |
| CadEval | 42% | — |
| AlgoTune | — | 1.85 |
Agentic & Tool Use GPT-5.4 leads
GPT-4.1: 34.4 (#40), GPT-5.4: 46.5 (#12)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| Terminal-Bench | — | 81.8% |
| APEX-Agents | — | 52.4% |
| Berkeley Function Calling Leaderboard | 54% | — |
| τ²-bench Banking | — | 39.4% |
| DeepResearch Bench | — | 35.1% |
| PostTrainBench | — | 19% |
| GBAEval | — | 45.1% |
| LMArena Search | — | 1197 |
| METR Time Horizons | — | 74.3% |
| Vending-Bench 2 | — | 6,144 |
Reasoning GPT-5.4 leads
GPT-4.1: 11.7 (#334), GPT-5.4: 61.7 (#17)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| ARC-AGI-2 | 0.4% | 74% |
| Kagi LLM Benchmark | 52.3% | 63.8% |
| ARC-AGI-1 | 5.5% | 93.7% |
| Chess Puzzles | 6% | 44% |
| EnigmaEval | 2.2% | 16% |
| LMArena Hard Prompts | 1384 | 1485 |
| DTBench | 68.3% | 94.4% |
| LMCA | 25.6% | 52% |
| Epoch Capabilities Index | 136.79 | 156.92 |
| ForecastBench | 61.5 | 59.5 |
| SimpleBench | 27% | — |
| NYT Connections (extended) | — | 91.3% |
| CritPt | — | 23.4% |
| Thematic Generalization | — | 80% |
| EBR-Bench | — | 25.4% |
| Mystery Game Puzzles | — | 37% |
Math GPT-5.4 leads
GPT-4.1: 21.4 (#269), GPT-5.4: 73.3 (#14)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 6% | 78.6% |
| OTIS Mock AIME 2024-2025 | 38.3% | 97.8% |
| LMArena Math | 1369 | 1488 |
| FrontierMath (Feb 2025 set) | 5.5% | 47.6% |
| FrontierMath Tier 4 (v1) | 0% | 27.1% |
| FrontierMath Tier 4 | — | 49% |
| MathArena Final-Answer Competitions | — | 83.1% |
| ProofBench | — | 56% |
| Omni-MATH | 47.1% | — |
| MATH Level 5 | 83% | — |
Knowledge GPT-5.4 leads
GPT-4.1: 37.0 (#152), GPT-5.4: 65.3 (#13)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 66.9% | 93.3% |
| Humanity's Last Exam | 5.4% | 36.2% |
| SimpleQA Verified | 31.1% | 45.1% |
| Vectara Hallucination Rate | 5.6% | 7% |
| LMArena Expert | 1364 | 1507 |
| MMLU-Pro | 81.1% | — |
| GPQA (HELM) | 65.9% | — |
Multimodal GPT-5.4 leads
GPT-4.1: 37.5 (#66), GPT-5.4: 43.8 (#19)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| LMArena Vision | 1210 | 1302 |
| GeoBench | 72% | — |
| Blueprint-Bench 2 | — | 27.1% |
| Furniture Assembly | — | 37.5% |
| LMArena Document | — | 1471 |
Multilingual GPT-5.4 leads
GPT-4.1: 49.6 (#128), GPT-5.4: 56.4 (#22)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1370 | 1465 |
| LMArena Chinese | 1382 | 1520 |
| LMArena French | 1383 | 1493 |
| LMArena German | 1380 | 1472 |
| LMArena Japanese | 1320 | 1484 |
| LMArena Korean | 1339 | 1449 |
| LMArena Russian | 1377 | 1480 |
| LMArena Spanish | 1375 | 1453 |
Instruction Following GPT-5.4 leads
GPT-4.1: 71.5 (#148), GPT-5.4: 77.3 (#25)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1367 | 1469 |
| IFEval | 83.8% | — |
Long Context GPT-5.4 leads
GPT-4.1: 40.0 (#158), GPT-5.4: 50.3 (#8)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1385 | 1474 |
| Fiction.LiveBench | 63.9% | — |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |
Writing & Preference GPT-5.4 leads
GPT-4.1: 57.8 (#120), GPT-5.4: 72.1 (#18)
| Benchmark | GPT-4.1 | GPT-5.4 |
|---|---|---|
| LMArena Text | 1383 | 1469 |
| LMArena Creative Writing | 1363 | 1439 |
| EQ-Bench Creative Writing | 1420 | 1840 |
| LMArena Multi-Turn | 1398 | 1482 |
| WildBench | 85.4% | — |
| EQ-Bench 4 | — | 1272 |
Frequently asked questions
Is GPT-4.1 better than GPT-5.4?
GPT-5.4 is the stronger model overall, scoring 59.5 to 35.7 on the Noometry Index. GPT-4.1 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-4.1 or GPT-5.4?
GPT-4.1 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-4.1 or GPT-5.4 better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.8 versus 34.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 1.05M.
How many benchmarks do GPT-4.1 and GPT-5.4 share?
39 benchmarks have published results for both models. GPT-4.1 has 52 scored results on Noometry and GPT-5.4 has 68.