# GPT-5.2 vs GPT-5.6 Sol

> GPT-5.6 Sol is the stronger model overall, scoring 64.4 to 54.1 on the Noometry Index. GPT-5.2 costs 1.7× less per token, which makes it the better buy when GPT-5.6 Sol's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-5-2-vs-gpt-5-6-sol
- Last updated: 2026-10-10
- Shared benchmarks: 46

## Summary

- They share 46 benchmarks with published results for both. GPT-5.2 scores higher in 1 category and GPT-5.6 Sol in 9 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Sol leads 85.5 to 59.8.
- The biggest single-benchmark swing is ProofBench: 15% for GPT-5.2 and 83% for GPT-5.6 Sol.
- GPT-5.2 is cheaper at $1.75 / $14 per million input/output tokens, against $4 / $20 for GPT-5.6 Sol.
- GPT-5.6 Sol accepts more context: 1.05M tokens versus 400K.

## Snapshot

| | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 54.1 | 64.4 |
| Rank | 31 | 6 |
| Context | 400K | 1.05M |
| Input $/M | $1.75 | $4 |
| Output $/M | $14 | $20 |
| Weights | Proprietary | Proprietary |

## Coding

- GPT-5.2: 52.2 (#32)
- GPT-5.6 Sol: 61.1 (#7)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| LMArena WebDev | 1415 | 1619 |
| WeirdML | 72.2% | 89.4% |
| LMArena Coding | 1447 | 1493 |
| ALE-Bench | 1,294 | 2,177 |
| SWE-bench Verified | 73.8% | — |
| DeepSWE | — | 72.7% |
| FrontierCode | — | 47.5% |
| SWE-bench Verified (bash only) | 72.8% | — |
| CursorBench | — | 41.7% |
| SWE-bench Multilingual | 66.7% | — |
| FrontierSWE | — | 32.2% |
| SciCode | — | 57.1% |
| GSO | 27.4% | — |
| MirrorCode | — | 20% |
| AlgoTune | 2.05 | — |

## Agentic & Tool Use

- GPT-5.2: 40.0 (#24)
- GPT-5.6 Sol: 50.2 (#7)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| τ²-bench Banking | 32.2% | 46.9% |
| LMArena Search | 1207 | 1257 |
| Vending-Bench 2 | 3,591 | 9,619 |
| Terminal-Bench | 64.9% | — |
| APEX-Agents | — | 51.4% |
| Berkeley Function Calling Leaderboard | 55.9% | — |
| OSWorld 2.0 | — | 27.3% |
| 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 | — | 36.2% |
| BALROG | — | 60% |
| GBAEval | — | 52.6% |
| GDP.pdf | — | 30.7% |
| METR Time Horizons | 75.3% | — |

## Reasoning

- GPT-5.2: 50.1 (#33)
- GPT-5.6 Sol: 74.7 (#6)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| ARC-AGI-2 | 52.9% | 92.5% |
| SimpleBench | 45.8% | 71.7% |
| Kagi LLM Benchmark | 73.3% | 67% |
| NYT Connections (extended) | 83.6% | 93.8% |
| ARC-AGI-1 | 86.2% | 97.5% |
| Chess Puzzles | 49% | 64% |
| EnigmaEval | 10.4% | 37.1% |
| EBR-Bench | 23% | 44.8% |
| LMArena Hard Prompts | 1445 | 1481 |
| Mystery Game Puzzles | 23% | 58% |
| DTBench | 90.9% | 96% |
| LMCA | 43.9% | 59.2% |
| Epoch Capabilities Index | 153.45 | 161.99 |
| CritPt | — | 32.3% |
| Surface Evolver Bench | — | 93.1% |
| Bench to the Future 3 | — | 0.14 |
| ForecastBench | 60.1 | — |

## Math

- GPT-5.2: 59.8 (#32)
- GPT-5.6 Sol: 85.5 (#7)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| FrontierMath (Tiers 1-3) | 67.4% | 89.1% |
| FrontierMath Tier 4 | 31.7% | 82.9% |
| OTIS Mock AIME 2024-2025 | 96.1% | 100% |
| ProofBench | 15% | 83% |
| LMArena Math | 1439 | 1472 |
| MathArena Final-Answer Competitions | 72% | — |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Erdős | — | 0% |
| FrontierMath Tier 4 (v1) | 18.8% | — |

## Knowledge

- GPT-5.2: 59.2 (#29)
- GPT-5.6 Sol: 64.3 (#16)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| GPQA Diamond | 91.4% | 93.5% |
| SimpleQA Verified | 37.1% | 69.7% |
| Vectara Hallucination Rate | 8.4% | 12.4% |
| LMArena Expert | 1445 | 1517 |
| Humanity's Last Exam | 27.8% | — |

## Multimodal

- GPT-5.2: 50.9 (#7)
- GPT-5.6 Sol: 48.9 (#8)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| LMArena Vision | 1268 | 1281 |
| Furniture Assembly | 38.3% | 56.7% |
| LMArena Document | 1405 | 1483 |
| VPCT | 84% | — |
| Blueprint-Bench 2 | — | 33.6% |

## Multilingual

- GPT-5.2: 53.6 (#63)
- GPT-5.6 Sol: 55.4 (#32)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| LMArena Non-English | 1425 | 1450 |
| LMArena Chinese | 1460 | 1523 |
| LMArena French | 1456 | 1475 |
| LMArena German | 1448 | 1476 |
| LMArena Japanese | 1421 | 1475 |
| LMArena Korean | 1390 | 1434 |
| LMArena Russian | 1440 | 1466 |
| LMArena Spanish | 1433 | 1449 |

## Instruction Following

- GPT-5.2: 74.9 (#83)
- GPT-5.6 Sol: 77.8 (#16)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| LMArena Instruction Following | 1417 | 1480 |

## Long Context

- GPT-5.2: 44.0 (#74)
- GPT-5.6 Sol: 45.3 (#42)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| LMArena Longer Query | 1428 | 1479 |
| CL-bench | 18.2% | — |

## Writing & Preference

- GPT-5.2: 67.0 (#30)
- GPT-5.6 Sol: 73.5 (#12)

| Benchmark | GPT-5.2 | GPT-5.6 Sol |
|---|---|---|
| LMArena Text | 1439 | 1456 |
| LMArena Creative Writing | 1402 | 1450 |
| EQ-Bench Creative Writing | 1703 | 1972 |
| LMArena Multi-Turn | 1458 | 1456 |
| EQ-Bench 4 | — | 1250 |

## FAQ

### Is GPT-5.2 better than GPT-5.6 Sol?

GPT-5.6 Sol is the stronger model overall, scoring 64.4 to 54.1 on the Noometry Index. GPT-5.2 costs 1.7× less per token, which makes it the better buy when GPT-5.6 Sol's lead doesn't matter for your workload.

### Which is cheaper, GPT-5.2 or GPT-5.6 Sol?

GPT-5.2 is cheaper. It lists at $1.75 per million input tokens and $14 per million output tokens; GPT-5.6 Sol lists at $4 and $20.

### Is GPT-5.2 or GPT-5.6 Sol better for coding?

GPT-5.6 Sol scores higher on coding benchmarks: 61.1 versus 52.2 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.6 Sol does, with 1.05M tokens against 400K.

### How many benchmarks do GPT-5.2 and GPT-5.6 Sol share?

46 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and GPT-5.6 Sol has 64.