Model comparison
GPT-5.2 vs o4-mini
GPT-5.2 is the stronger model overall, scoring 54.1 to 41.6 on the Noometry Index. o4-mini costs 2.5× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Last verified . 47 shared benchmarks.
Summary
- They share 47 benchmarks with published results for both. GPT-5.2 scores higher in 8 categories and o4-mini in 2 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 24.6.
- The biggest single-benchmark swing is ARC-AGI-2: 52.9% for GPT-5.2 and 6.1% for o4-mini.
- o4-mini is cheaper at $1.10 / $4.40 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 200K.
Side by side
| GPT-5.2 | o4-mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 54.1 | 41.6 |
| Released | 2025-12-11 | 2025-04-16 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 200K |
| Max output | 128K | 100K |
| Input $ / M tokens | $1.75 | $1.10 |
| Output $ / M tokens | $14 | $4.40 |
| Results tracked | 67 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GPT-5.2: 51.6 (#37), o4-mini: 40.9 (#127)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| SWE-bench Verified (bash only) | 72.8% | 45% |
| GSO | 27.4% | 3.6% |
| WeirdML | 72.2% | 52.6% |
| LMArena Coding | 1447 | 1368 |
| ALE-Bench | 1,294 | 826.17 |
| AlgoTune | 2.05 | 1.72 |
| SWE-bench Verified | 73.8% | — |
| Aider Polyglot | — | 72% |
| LMArena WebDev | 1416 | — |
| SWE-bench Multilingual | 66.7% | — |
| CadEval | — | 62% |
Agentic & Tool Use GPT-5.2 leads
GPT-5.2: 40.2 (#24), o4-mini: 32.6 (#61)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| Berkeley Function Calling Leaderboard | 55.9% | 53.2% |
| GDPval | 49.7% | 25.3% |
| METR Time Horizons | 75.3% | 63.9% |
| Terminal-Bench | 64.9% | — |
| 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 | — |
| Vending-Bench 2 | 3,591 | — |
Reasoning GPT-5.2 leads
GPT-5.2: 50.2 (#35), o4-mini: 24.6 (#162)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| ARC-AGI-2 | 52.9% | 6.1% |
| SimpleBench | 45.8% | 38.7% |
| Kagi LLM Benchmark | 73.3% | 67.6% |
| ARC-AGI-1 | 86.2% | 58.7% |
| Chess Puzzles | 49% | 26% |
| EnigmaEval | 10.4% | 9.2% |
| LMArena Hard Prompts | 1445 | 1351 |
| Mystery Game Puzzles | 23% | 5% |
| DTBench | 90.9% | 77.6% |
| LMCA | 43.9% | 26.5% |
| Epoch Capabilities Index | 153.45 | 145.64 |
| ForecastBench | 60.1 | 61.8 |
| NYT Connections (extended) | 83.6% | — |
| CritPt | — | 0.6% |
| EBR-Bench | 23% | — |
Math GPT-5.2 leads
GPT-5.2: 60.0 (#38), o4-mini: 40.8 (#89)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 67.4% | 36.1% |
| FrontierMath Tier 4 | 31.7% | 4.9% |
| OTIS Mock AIME 2024-2025 | 96.1% | 81.7% |
| LMArena Math | 1440 | 1389 |
| FrontierMath (Feb 2025 set) | 40.7% | 24.8% |
| FrontierMath Tier 4 (v1) | 18.8% | 6.3% |
| MathArena Final-Answer Competitions | 72% | — |
| ProofBench | 15% | — |
| Omni-MATH | — | 72% |
| MATH Level 5 | — | 97.8% |
Knowledge GPT-5.2 leads
GPT-5.2: 59.3 (#32), o4-mini: 43.6 (#91)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| GPQA Diamond | 91.4% | 79.6% |
| Humanity's Last Exam | 27.8% | 18.1% |
| SimpleQA Verified | 37.1% | 19.6% |
| Vectara Hallucination Rate | 8.4% | 18.6% |
| LMArena Expert | 1445 | 1343 |
| MMLU-Pro | — | 82% |
| Confabulations | — | 15.8% |
| GPQA (HELM) | — | 73.5% |
Multimodal GPT-5.2 leads
GPT-5.2: 51.3 (#7), o4-mini: 40.2 (#49)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| LMArena Vision | 1268 | 1194 |
| VPCT | 84% | 57.5% |
| GeoBench | — | 64% |
| Furniture Assembly | 38.3% | — |
| LMArena Document | 1405 | — |
Multilingual GPT-5.2 leads
GPT-5.2: 53.4 (#67), o4-mini: 47.0 (#154)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| LMArena Non-English | 1425 | 1337 |
| LMArena Chinese | 1460 | 1354 |
| LMArena French | 1455 | 1364 |
| LMArena German | 1448 | 1336 |
| LMArena Japanese | 1420 | 1308 |
| LMArena Korean | 1392 | 1312 |
| LMArena Russian | 1440 | 1334 |
| LMArena Spanish | 1433 | 1347 |
Instruction Following Too close to call
GPT-5.2: 74.7 (#89), o4-mini: 75.2 (#68)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| LMArena Instruction Following | 1417 | 1321 |
| IFEval | — | 92.8% |
Long Context o4-mini leads
GPT-5.2: 44.0 (#78), o4-mini: 45.5 (#33)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| LMArena Longer Query | 1428 | 1315 |
| Fiction.LiveBench | — | 77.8% |
| CL-bench | 18.2% | — |
Writing & Preference GPT-5.2 leads
GPT-5.2: 66.8 (#32), o4-mini: 54.0 (#152)
| Benchmark | GPT-5.2 | o4-mini |
|---|---|---|
| LMArena Text | 1439 | 1353 |
| LMArena Creative Writing | 1401 | 1294 |
| LMArena Multi-Turn | 1458 | 1350 |
| Short-Story Creative Writing | — | 75% |
| EQ-Bench Creative Writing | 1703 | — |
| WildBench | — | 85.4% |
Frequently asked questions
Is GPT-5.2 better than o4-mini?
GPT-5.2 is the stronger model overall, scoring 54.1 to 41.6 on the Noometry Index. o4-mini costs 2.5× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Which is cheaper, GPT-5.2 or o4-mini?
o4-mini is cheaper. It lists at $1.10 per million input tokens and $4.40 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GPT-5.2 or o4-mini better for coding?
GPT-5.2 scores higher on coding benchmarks: 51.6 versus 40.9 in the Noometry coding category.
Which has the bigger context window?
GPT-5.2 does, with 400K tokens against 200K.
How many benchmarks do GPT-5.2 and o4-mini share?
47 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and o4-mini has 60.