Model comparison
GPT-4 Turbo vs GPT-5.2
GPT-5.2 is the stronger model overall, scoring 54.1 to 30.5 on the Noometry Index.
Last verified . 29 shared benchmarks.
Summary
- They share 29 benchmarks with published results for both. GPT-4 Turbo scores higher in 0 categories and GPT-5.2 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.2 leads 60.0 to 9.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 6.7% for GPT-4 Turbo and 96.1% for GPT-5.2.
- GPT-5.2 is cheaper at $1.75 / $14 per million input/output tokens, against $10 / $30 for GPT-4 Turbo.
- GPT-5.2 accepts more context: 400K tokens versus 128K.
Side by side
| GPT-4 Turbo | GPT-5.2 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 30.5 | 54.1 |
| Released | 2023-11-06 | 2025-12-11 |
| Weights | Proprietary | Proprietary |
| Context window | 128K | 400K |
| Max output | 4K | 128K |
| Input $ / M tokens | $10 | $1.75 |
| Output $ / M tokens | $30 | $14 |
| Results tracked | 36 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GPT-4 Turbo: 33.8 (#249), GPT-5.2: 51.6 (#37)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| WeirdML | 18% | 72.2% |
| LMArena Coding | 1268 | 1447 |
| SWE-bench Verified | — | 73.8% |
| SWE-bench Verified (bash only) | — | 72.8% |
| LMArena WebDev | — | 1416 |
| SWE-bench Multilingual | — | 66.7% |
| GSO | — | 27.4% |
| BigCodeBench Instruct | 48.2% | — |
| BigCodeBench Complete | 58.2% | — |
| ALE-Bench | — | 1,294 |
| AlgoTune | — | 2.05 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73.3% | — |
Agentic & Tool Use Not comparable
GPT-4 Turbo: —, GPT-5.2: 40.2 (#24)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| METR Time Horizons | 36.7% | 75.3% |
| Terminal-Bench | — | 64.9% |
| Berkeley Function Calling Leaderboard | — | 55.9% |
| 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 |
| Vending-Bench 2 | — | 3,591 |
Reasoning GPT-5.2 leads
GPT-4 Turbo: 15.3 (#317), GPT-5.2: 50.2 (#35)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| SimpleBench | 25.1% | 45.8% |
| Chess Puzzles | 6% | 49% |
| LMArena Hard Prompts | 1251 | 1445 |
| DTBench | 61.6% | 90.9% |
| LMCA | 9.8% | 43.9% |
| Epoch Capabilities Index | 127.25 | 153.45 |
| ForecastBench | 59.4 | 60.1 |
| ARC-AGI-2 | — | 52.9% |
| Kagi LLM Benchmark | — | 73.3% |
| NYT Connections (extended) | — | 83.6% |
| ARC-AGI-1 | — | 86.2% |
| EnigmaEval | — | 10.4% |
| EBR-Bench | — | 23% |
| Mystery Game Puzzles | — | 23% |
Math GPT-5.2 leads
GPT-4 Turbo: 9.0 (#322), GPT-5.2: 60.0 (#38)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 0.7% | 67.4% |
| OTIS Mock AIME 2024-2025 | 6.7% | 96.1% |
| LMArena Math | 1272 | 1440 |
| FrontierMath Tier 4 | — | 31.7% |
| MathArena Final-Answer Competitions | — | 72% |
| ProofBench | — | 15% |
| MATH Level 5 | 46.7% | — |
| FrontierMath (Feb 2025 set) | — | 40.7% |
| FrontierMath Tier 4 (v1) | — | 18.8% |
Knowledge GPT-5.2 leads
GPT-4 Turbo: 24.3 (#268), GPT-5.2: 59.3 (#32)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 46.6% | 91.4% |
| LMArena Expert | 1223 | 1445 |
| Humanity's Last Exam | — | 27.8% |
| SimpleQA Verified | — | 37.1% |
| Confabulations | 28.4% | — |
| Vectara Hallucination Rate | — | 8.4% |
| MMLU | 81.3% | — |
Multimodal GPT-5.2 leads
GPT-4 Turbo: 30.6 (#110), GPT-5.2: 51.3 (#7)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| LMArena Vision | 1090 | 1268 |
| VPCT | — | 84% |
| Furniture Assembly | — | 38.3% |
| LMArena Document | — | 1405 |
Multilingual GPT-5.2 leads
GPT-4 Turbo: 40.5 (#216), GPT-5.2: 53.4 (#67)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1245 | 1425 |
| LMArena Chinese | 1242 | 1460 |
| LMArena French | 1276 | 1455 |
| LMArena German | 1259 | 1448 |
| LMArena Japanese | 1194 | 1420 |
| LMArena Korean | 1187 | 1392 |
| LMArena Russian | 1259 | 1440 |
| LMArena Spanish | 1260 | 1433 |
Instruction Following GPT-5.2 leads
GPT-4 Turbo: 65.8 (#216), GPT-5.2: 74.7 (#89)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1249 | 1417 |
Long Context GPT-5.2 leads
GPT-4 Turbo: 38.0 (#206), GPT-5.2: 44.0 (#78)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| LMArena Longer Query | 1254 | 1428 |
| CL-bench | — | 18.2% |
Writing & Preference GPT-5.2 leads
GPT-4 Turbo: 47.7 (#206), GPT-5.2: 66.8 (#32)
| Benchmark | GPT-4 Turbo | GPT-5.2 |
|---|---|---|
| LMArena Text | 1272 | 1439 |
| LMArena Creative Writing | 1269 | 1401 |
| LMArena Multi-Turn | 1267 | 1458 |
| EQ-Bench Creative Writing | — | 1703 |
Frequently asked questions
Is GPT-4 Turbo better than GPT-5.2?
GPT-5.2 is the stronger model overall, scoring 54.1 to 30.5 on the Noometry Index.
Which is cheaper, GPT-4 Turbo or GPT-5.2?
GPT-5.2 is cheaper. It lists at $1.75 per million input tokens and $14 per million output tokens; GPT-4 Turbo lists at $10 and $30.
Is GPT-4 Turbo or GPT-5.2 better for coding?
GPT-5.2 scores higher on coding benchmarks: 51.6 versus 33.8 in the Noometry coding category.
Which has the bigger context window?
GPT-5.2 does, with 400K tokens against 128K.
How many benchmarks do GPT-4 Turbo and GPT-5.2 share?
29 benchmarks have published results for both models. GPT-4 Turbo has 36 scored results on Noometry and GPT-5.2 has 67.