# GPT-5.2 vs GPT-5.5

> GPT-5.5 is the stronger model overall, scoring 63.4 to 54.1 on the Noometry Index. GPT-5.2 costs 2.3× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-5-2-vs-gpt-5-5
- Last updated: 2026-10-10
- Shared benchmarks: 54

## Summary

- They share 54 benchmarks with published results for both. GPT-5.2 scores higher in 1 category and GPT-5.5 in 9 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.5 leads 72.7 to 50.1.
- The biggest single-benchmark swing is FrontierMath Tier 4: 31.7% for GPT-5.2 and 72.5% for GPT-5.5.
- GPT-5.2 is cheaper at $1.75 / $14 per million input/output tokens, against $5 / $30 for GPT-5.5.
- GPT-5.5 accepts more context: 1.05M tokens versus 400K.

## Snapshot

| | GPT-5.2 | GPT-5.5 |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 54.1 | 63.4 |
| Rank | 31 | 8 |
| Context | 400K | 1.05M |
| Input $/M | $1.75 | $5 |
| Output $/M | $14 | $30 |
| Weights | Proprietary | Proprietary |

## Coding

- GPT-5.2: 52.2 (#32)
- GPT-5.5: 58.3 (#13)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| SWE-bench Verified | 73.8% | 80.6% |
| LMArena WebDev | 1415 | 1512 |
| GSO | 27.4% | 40.2% |
| WeirdML | 72.2% | 84.9% |
| LMArena Coding | 1447 | 1494 |
| ALE-Bench | 1,294 | 1,943 |
| DeepSWE | — | 67% |
| FrontierCode | — | 43% |
| SWE-bench Verified (bash only) | 72.8% | — |
| SWE-bench Multilingual | 66.7% | — |
| SciCode | — | 56.1% |
| MirrorCode | — | 10% |
| AlgoTune | 2.05 | — |

## Agentic & Tool Use

- GPT-5.2: 40.0 (#24)
- GPT-5.5: 50.7 (#6)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| Terminal-Bench | 64.9% | 84.7% |
| Remote Labor Index | 2.5% | 6.3% |
| τ²-bench Banking | 32.2% | 44.6% |
| DeepResearch Bench | 41.1% | 54% |
| LMArena Search | 1207 | 1242 |
| Vending-Bench 2 | 3,591 | 7,524 |
| APEX-Agents | — | 55.1% |
| Berkeley Function Calling Leaderboard | 55.9% | — |
| OSWorld 2.0 | — | 13% |
| GDPval | 49.7% | — |
| τ²-bench Airline | 83% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| PostTrainBench | — | 27.2% |
| ExploitBench | — | 47.4% |
| GBAEval | — | 53.2% |
| GDP.pdf | — | 26% |
| METR Time Horizons | 75.3% | — |

## Reasoning

- GPT-5.2: 50.1 (#33)
- GPT-5.5: 72.7 (#8)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| ARC-AGI-2 | 52.9% | 85% |
| SimpleBench | 45.8% | 69% |
| Kagi LLM Benchmark | 73.3% | 88.8% |
| NYT Connections (extended) | 83.6% | 96.2% |
| ARC-AGI-1 | 86.2% | 95% |
| Chess Puzzles | 49% | 54% |
| EBR-Bench | 23% | 34.3% |
| LMArena Hard Prompts | 1445 | 1489 |
| Mystery Game Puzzles | 23% | 56% |
| DTBench | 90.9% | 96% |
| LMCA | 43.9% | 54.3% |
| Epoch Capabilities Index | 153.45 | 159.25 |
| ForecastBench | 60.1 | 60.6 |
| CritPt | — | 27.1% |
| EnigmaEval | 10.4% | — |
| Surface Evolver Bench | — | 88.1% |
| Bench to the Future 3 | — | 0.14 |

## Math

- GPT-5.2: 59.8 (#32)
- GPT-5.5: 81.5 (#9)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 67.4% | 85.3% |
| FrontierMath Tier 4 | 31.7% | 72.5% |
| MathArena Final-Answer Competitions | 72% | 94.3% |
| OTIS Mock AIME 2024-2025 | 96.1% | 100% |
| ProofBench | 15% | 50% |
| LMArena Math | 1439 | 1486 |
| FrontierMath (Feb 2025 set) | 40.7% | 51.7% |
| FrontierMath Tier 4 (v1) | 18.8% | 35.4% |
| FrontierMath Erdős | — | 0% |

## Knowledge

- GPT-5.2: 59.2 (#29)
- GPT-5.5: 64.4 (#14)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| GPQA Diamond | 91.4% | 94% |
| SimpleQA Verified | 37.1% | 63% |
| Vectara Hallucination Rate | 8.4% | 9.3% |
| LMArena Expert | 1445 | 1508 |
| Humanity's Last Exam | 27.8% | — |

## Multimodal

- GPT-5.2: 50.9 (#7)
- GPT-5.5: 47.1 (#11)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| LMArena Vision | 1268 | 1297 |
| Furniture Assembly | 38.3% | 44.2% |
| LMArena Document | 1405 | 1486 |
| VPCT | 84% | — |
| Blueprint-Bench 2 | — | 36.2% |

## Multilingual

- GPT-5.2: 53.6 (#63)
- GPT-5.5: 56.5 (#19)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| LMArena Non-English | 1425 | 1467 |
| LMArena Chinese | 1460 | 1535 |
| LMArena French | 1456 | 1486 |
| LMArena German | 1448 | 1478 |
| LMArena Japanese | 1421 | 1498 |
| LMArena Korean | 1390 | 1461 |
| LMArena Russian | 1440 | 1473 |
| LMArena Spanish | 1433 | 1469 |

## Instruction Following

- GPT-5.2: 74.9 (#83)
- GPT-5.5: 77.7 (#18)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| LMArena Instruction Following | 1417 | 1479 |

## Long Context

- GPT-5.2: 44.0 (#74)
- GPT-5.5: 48.3 (#12)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| LMArena Longer Query | 1428 | 1484 |
| CL-bench | 18.2% | — |
| CL-bench Life | — | 22.2% |

## Writing & Preference

- GPT-5.2: 67.0 (#30)
- GPT-5.5: 72.8 (#13)

| Benchmark | GPT-5.2 | GPT-5.5 |
|---|---|---|
| LMArena Text | 1439 | 1471 |
| LMArena Creative Writing | 1402 | 1455 |
| EQ-Bench Creative Writing | 1703 | 1844 |
| LMArena Multi-Turn | 1458 | 1477 |
| EQ-Bench 4 | — | 1315 |

## FAQ

### Is GPT-5.2 better than GPT-5.5?

GPT-5.5 is the stronger model overall, scoring 63.4 to 54.1 on the Noometry Index. GPT-5.2 costs 2.3× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.

### Which is cheaper, GPT-5.2 or GPT-5.5?

GPT-5.2 is cheaper. It lists at $1.75 per million input tokens and $14 per million output tokens; GPT-5.5 lists at $5 and $30.

### Is GPT-5.2 or GPT-5.5 better for coding?

GPT-5.5 scores higher on coding benchmarks: 58.3 versus 52.2 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.5 does, with 1.05M tokens against 400K.

### How many benchmarks do GPT-5.2 and GPT-5.5 share?

54 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and GPT-5.5 has 71.