# Claude Opus 5 vs GPT-5.2

> Claude Opus 5 is the stronger model overall, scoring 67.8 to 54.1 on the Noometry Index. GPT-5.2 costs 2.1× less per token, which makes it the better buy when Claude Opus 5's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/claude-opus-5-vs-gpt-5-2
- Last updated: 2026-10-10
- Shared benchmarks: 42

## Summary

- They share 42 benchmarks with published results for both. Claude Opus 5 scores higher in 9 categories and GPT-5.2 in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Claude Opus 5 leads 77.2 to 50.2.
- The biggest single-benchmark swing is ProofBench: 99% for Claude Opus 5 and 15% for GPT-5.2.
- GPT-5.2 is cheaper at $1.75 / $14 per million input/output tokens, against $5 / $25 for Claude Opus 5.
- Claude Opus 5 accepts more context: 1M tokens versus 400K.

## Snapshot

| | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 67.8 | 54.1 |
| Rank | 4 | 34 |
| Context | 1M | 400K |
| Input $/M | $5 | $1.75 |
| Output $/M | $25 | $14 |
| Weights | Proprietary | Proprietary |

## Coding

- Claude Opus 5: 67.5 (#5)
- GPT-5.2: 51.6 (#37)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| LMArena WebDev | 1691 | 1416 |
| WeirdML | 91.8% | 72.2% |
| LMArena Coding | 1534 | 1447 |
| ALE-Bench | 2,165 | 1,294 |
| SWE-bench Verified | — | 73.8% |
| DeepSWE | 73.6% | — |
| FrontierCode | 53.4% | — |
| SWE-bench Verified (bash only) | — | 72.8% |
| CursorBench | 46.6% | — |
| SWE-bench Multilingual | — | 66.7% |
| FrontierSWE | 52% | — |
| SciCode | 56.4% | — |
| GSO | — | 27.4% |
| AlgoTune | — | 2.05 |

## Agentic & Tool Use

- Claude Opus 5: 55.6 (#1)
- GPT-5.2: 40.2 (#24)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| τ²-bench Banking | 48.7% | 32.2% |
| Vending-Bench 2 | 11,182 | 3,591 |
| Terminal-Bench | — | 64.9% |
| APEX-Agents | 65.8% | — |
| Berkeley Function Calling Leaderboard | — | 55.9% |
| OSWorld 2.0 | 31.4% | — |
| GDPval | — | 49.7% |
| Remote Labor Index | — | 2.5% |
| τ²-bench Airline | — | 83% |
| τ²-bench Retail | — | 81.6% |
| τ²-bench Telecom | — | 89.7% |
| DeepResearch Bench | — | 41.1% |
| PostTrainBench | 35% | — |
| BALROG | 63.4% | — |
| GBAEval | 79.6% | — |
| GDP.pdf | 24% | — |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 75.3% |

## Reasoning

- Claude Opus 5: 77.2 (#4)
- GPT-5.2: 50.2 (#35)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| ARC-AGI-2 | 90.4% | 52.9% |
| SimpleBench | 80.6% | 45.8% |
| NYT Connections (extended) | 94.3% | 83.6% |
| ARC-AGI-1 | 97.5% | 86.2% |
| Chess Puzzles | 42% | 49% |
| EBR-Bench | 45.7% | 23% |
| LMArena Hard Prompts | 1526 | 1445 |
| Mystery Game Puzzles | 59% | 23% |
| DTBench | 97.9% | 90.9% |
| LMCA | 64.5% | 43.9% |
| Epoch Capabilities Index | 162.78 | 153.45 |
| Kagi LLM Benchmark | — | 73.3% |
| CritPt | 29.1% | — |
| EnigmaEval | — | 10.4% |
| Bench to the Future 3 | 0.12 | — |
| ForecastBench | — | 60.1 |

## Math

- Claude Opus 5: 86.2 (#8)
- GPT-5.2: 60.0 (#38)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 85.6% | 67.4% |
| FrontierMath Tier 4 | 73.2% | 31.7% |
| OTIS Mock AIME 2024-2025 | 98.9% | 96.1% |
| ProofBench | 99% | 15% |
| LMArena Math | 1531 | 1440 |
| MathArena Final-Answer Competitions | — | 72% |
| FrontierMath (Feb 2025 set) | — | 40.7% |
| FrontierMath Tier 4 (v1) | — | 18.8% |

## Knowledge

- Claude Opus 5: 66.8 (#9)
- GPT-5.2: 59.3 (#32)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 93.9% | 91.4% |
| SimpleQA Verified | 59.9% | 37.1% |
| LMArena Expert | 1557 | 1445 |
| Humanity's Last Exam | — | 27.8% |
| Vectara Hallucination Rate | — | 8.4% |

## Multimodal

- Claude Opus 5: 50.8 (#8)
- GPT-5.2: 51.3 (#7)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| LMArena Vision | 1319 | 1268 |
| Furniture Assembly | 60.8% | 38.3% |
| LMArena Document | 1516 | 1405 |
| VPCT | — | 84% |
| Blueprint-Bench 2 | 30.4% | — |

## Multilingual

- Claude Opus 5: 58.8 (#4)
- GPT-5.2: 53.4 (#67)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1501 | 1425 |
| LMArena Chinese | 1574 | 1460 |
| LMArena French | 1519 | 1455 |
| LMArena German | 1524 | 1448 |
| LMArena Japanese | 1516 | 1420 |
| LMArena Korean | 1521 | 1392 |
| LMArena Russian | 1507 | 1440 |
| LMArena Spanish | 1519 | 1433 |

## Instruction Following

- Claude Opus 5: 79.2 (#7)
- GPT-5.2: 74.7 (#89)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1517 | 1417 |

## Long Context

- Claude Opus 5: 46.5 (#21)
- GPT-5.2: 44.0 (#78)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| LMArena Longer Query | 1515 | 1428 |
| CL-bench | — | 18.2% |

## Writing & Preference

- Claude Opus 5: 79.2 (#1)
- GPT-5.2: 66.8 (#32)

| Benchmark | Claude Opus 5 | GPT-5.2 |
|---|---|---|
| LMArena Text | 1507 | 1439 |
| LMArena Creative Writing | 1491 | 1401 |
| EQ-Bench Creative Writing | 2133 | 1703 |
| LMArena Multi-Turn | 1499 | 1458 |
| EQ-Bench 4 | 1385 | — |

## FAQ

### Is Claude Opus 5 better than GPT-5.2?

Claude Opus 5 is the stronger model overall, scoring 67.8 to 54.1 on the Noometry Index. GPT-5.2 costs 2.1× less per token, which makes it the better buy when Claude Opus 5's lead doesn't matter for your workload.

### Which is cheaper, Claude Opus 5 or GPT-5.2?

GPT-5.2 is cheaper. It lists at $1.75 per million input tokens and $14 per million output tokens; Claude Opus 5 lists at $5 and $25.

### Is Claude Opus 5 or GPT-5.2 better for coding?

Claude Opus 5 scores higher on coding benchmarks: 67.5 versus 51.6 in the Noometry coding category.

### Which has the bigger context window?

Claude Opus 5 does, with 1M tokens against 400K.

### How many benchmarks do Claude Opus 5 and GPT-5.2 share?

42 benchmarks have published results for both models. Claude Opus 5 has 57 scored results on Noometry and GPT-5.2 has 67.