Model comparison
GPT-4.1 vs GPT-5.2
GPT-5.2 is the stronger model overall, scoring 54.1 to 35.7 on the Noometry Index.
Last verified . 42 shared benchmarks.
Summary
- They share 42 benchmarks with published results for both. GPT-4.1 scores higher in 0 categories and GPT-5.2 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.1 to 11.7.
- The biggest single-benchmark swing is ARC-AGI-1: 5.5% for GPT-4.1 and 86.2% for GPT-5.2.
- GPT-4.1 is cheaper at $2 / $8 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-4.1 accepts more context: 1.05M tokens versus 400K.
Side by side
| GPT-4.1 | GPT-5.2 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 35.7 | 54.1 |
| Released | 2025-04-14 | 2025-12-11 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 400K |
| Max output | 33K | 128K |
| Input $ / M tokens | $2 | $1.75 |
| Output $ / M tokens | $8 | $14 |
| Results tracked | 52 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GPT-4.1: 34.2 (#228), GPT-5.2: 52.2 (#32)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| SWE-bench Verified | 48.5% | 73.8% |
| SWE-bench Verified (bash only) | 39.6% | 72.8% |
| WeirdML | 39% | 72.2% |
| LMArena Coding | 1391 | 1447 |
| ALE-Bench | 558.1 | 1,294 |
| Aider Polyglot | 52.4% | — |
| LMArena WebDev | — | 1415 |
| SWE-bench Multilingual | — | 66.7% |
| GSO | — | 27.4% |
| CadEval | 42% | — |
| AlgoTune | — | 2.05 |
Agentic & Tool Use GPT-5.2 leads
GPT-4.1: 34.4 (#40), GPT-5.2: 40.0 (#24)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| Berkeley Function Calling Leaderboard | 54% | 55.9% |
| Terminal-Bench | — | 64.9% |
| GDPval | — | 49.7% |
| Remote Labor Index | — | 2.5% |
| τ²-bench Airline | — | 83% |
| τ²-bench Banking | — | 32.2% |
| τ²-bench Retail | — | 81.6% |
| τ²-bench Telecom | — | 89.7% |
| DeepResearch Bench | — | 41.1% |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 75.3% |
| Vending-Bench 2 | — | 3,591 |
Reasoning GPT-5.2 leads
GPT-4.1: 11.7 (#334), GPT-5.2: 50.1 (#33)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| ARC-AGI-2 | 0.4% | 52.9% |
| SimpleBench | 27% | 45.8% |
| Kagi LLM Benchmark | 52.3% | 73.3% |
| ARC-AGI-1 | 5.5% | 86.2% |
| Chess Puzzles | 6% | 49% |
| EnigmaEval | 2.2% | 10.4% |
| LMArena Hard Prompts | 1384 | 1445 |
| DTBench | 68.3% | 90.9% |
| LMCA | 25.6% | 43.9% |
| Epoch Capabilities Index | 136.79 | 153.45 |
| ForecastBench | 61.5 | 60.1 |
| NYT Connections (extended) | — | 83.6% |
| EBR-Bench | — | 23% |
| Mystery Game Puzzles | — | 23% |
Math GPT-5.2 leads
GPT-4.1: 21.4 (#269), GPT-5.2: 59.8 (#32)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 6% | 67.4% |
| OTIS Mock AIME 2024-2025 | 38.3% | 96.1% |
| LMArena Math | 1369 | 1439 |
| FrontierMath (Feb 2025 set) | 5.5% | 40.7% |
| FrontierMath Tier 4 (v1) | 0% | 18.8% |
| FrontierMath Tier 4 | — | 31.7% |
| MathArena Final-Answer Competitions | — | 72% |
| ProofBench | — | 15% |
| Omni-MATH | 47.1% | — |
| MATH Level 5 | 83% | — |
Knowledge GPT-5.2 leads
GPT-4.1: 37.0 (#152), GPT-5.2: 59.2 (#29)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 66.9% | 91.4% |
| Humanity's Last Exam | 5.4% | 27.8% |
| SimpleQA Verified | 31.1% | 37.1% |
| Vectara Hallucination Rate | 5.6% | 8.4% |
| LMArena Expert | 1364 | 1445 |
| MMLU-Pro | 81.1% | — |
| GPQA (HELM) | 65.9% | — |
Multimodal GPT-5.2 leads
GPT-4.1: 37.5 (#66), GPT-5.2: 50.9 (#7)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| LMArena Vision | 1210 | 1268 |
| GeoBench | 72% | — |
| VPCT | — | 84% |
| Furniture Assembly | — | 38.3% |
| LMArena Document | — | 1405 |
Multilingual GPT-5.2 leads
GPT-4.1: 49.6 (#128), GPT-5.2: 53.6 (#63)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1370 | 1425 |
| LMArena Chinese | 1382 | 1460 |
| LMArena French | 1383 | 1456 |
| LMArena German | 1380 | 1448 |
| LMArena Japanese | 1320 | 1421 |
| LMArena Korean | 1339 | 1390 |
| LMArena Russian | 1377 | 1440 |
| LMArena Spanish | 1375 | 1433 |
Instruction Following GPT-5.2 leads
GPT-4.1: 71.5 (#148), GPT-5.2: 74.9 (#83)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1367 | 1417 |
| IFEval | 83.8% | — |
Long Context GPT-5.2 leads
GPT-4.1: 40.0 (#158), GPT-5.2: 44.0 (#74)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| LMArena Longer Query | 1385 | 1428 |
| Fiction.LiveBench | 63.9% | — |
| CL-bench | — | 18.2% |
Writing & Preference GPT-5.2 leads
GPT-4.1: 57.8 (#120), GPT-5.2: 67.0 (#30)
| Benchmark | GPT-4.1 | GPT-5.2 |
|---|---|---|
| LMArena Text | 1383 | 1439 |
| LMArena Creative Writing | 1363 | 1402 |
| EQ-Bench Creative Writing | 1420 | 1703 |
| LMArena Multi-Turn | 1398 | 1458 |
| WildBench | 85.4% | — |
Frequently asked questions
Is GPT-4.1 better than GPT-5.2?
GPT-5.2 is the stronger model overall, scoring 54.1 to 35.7 on the Noometry Index.
Which is cheaper, GPT-4.1 or GPT-5.2?
GPT-4.1 is cheaper. It lists at $2 per million input tokens and $8 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GPT-4.1 or GPT-5.2 better for coding?
GPT-5.2 scores higher on coding benchmarks: 52.2 versus 34.2 in the Noometry coding category.
Which has the bigger context window?
GPT-4.1 does, with 1.05M tokens against 400K.
How many benchmarks do GPT-4.1 and GPT-5.2 share?
42 benchmarks have published results for both models. GPT-4.1 has 52 scored results on Noometry and GPT-5.2 has 67.