# Claude Sonnet 4.5 vs GPT-4

> Claude Sonnet 4.5 is the stronger model overall, scoring 44.1 to 29.1 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-sonnet-4-5-vs-gpt-4
- Last updated: 2026-10-10
- Shared benchmarks: 29

## Summary

- They share 29 benchmarks with published results for both. Claude Sonnet 4.5 scores higher in 8 categories and GPT-4 in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Claude Sonnet 4.5 leads 66.5 to 34.9.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 77.8% for Claude Sonnet 4.5 and 1.1% for GPT-4.
- Claude Sonnet 4.5 is cheaper at $3 / $15 per million input/output tokens, against $30 / $60 for GPT-4.
- Claude Sonnet 4.5 accepts more context: 200K tokens versus 8K.

## Snapshot

| | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 44.1 | 29.1 |
| Rank | 81 | 316 |
| Context | 200K | 8K |
| Input $/M | $3 | $30 |
| Output $/M | $15 | $60 |
| Weights | Proprietary | Proprietary |

## Coding

- Claude Sonnet 4.5: 47.3 (#61)
- GPT-4: 31.6 (#283)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| WeirdML | 47.7% | 12.4% |
| LMArena Coding | 1489 | 1254 |
| SWE-bench Verified | 71.3% | — |
| SWE-bench Verified (bash only) | 71.4% | — |
| LMArena WebDev | 1393 | — |
| SWE-bench Multilingual | 67% | — |
| SciCode | 44.7% | — |
| GSO | 14.7% | — |
| BigCodeBench Instruct | — | 46% |
| BigCodeBench Complete | — | 57.2% |
| ALE-Bench | 796.15 | — |
| AlgoTune | 1.52 | — |
| HumanEval+ | — | 79.3% |

## Agentic & Tool Use

- Claude Sonnet 4.5: 38.3 (#32)
- GPT-4: —

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| METR Time Horizons | 67.4% | 36.1% |
| Terminal-Bench | 46.5% | — |
| Berkeley Function Calling Leaderboard | 73.2% | — |
| GDPval | 42.5% | — |
| Remote Labor Index | 2.1% | — |
| τ²-bench Airline | 72% | — |
| τ²-bench Banking | 25.3% | — |
| τ²-bench Retail | 72.4% | — |
| τ²-bench Telecom | 84.9% | — |
| Cybench | 60% | — |
| DeepResearch Bench | 52.6% | — |
| OSWorld | 62.9% | — |
| LMArena Search | 1159 | — |
| Vending-Bench 2 | 3,839 | — |

## Reasoning

- Claude Sonnet 4.5: 26.9 (#125)
- GPT-4: 17.8 (#289)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| Chess Puzzles | 12% | 4% |
| LMArena Hard Prompts | 1462 | 1241 |
| Mystery Game Puzzles | 17% | 12% |
| DTBench | 83.2% | 62.7% |
| LMCA | 38.8% | 17.1% |
| Epoch Capabilities Index | 146.84 | 125.89 |
| ForecastBench | 61.9 | 57.8 |
| ARC-AGI-2 | 13.6% | — |
| SimpleBench | 54.3% | — |
| Kagi LLM Benchmark | 57.9% | — |
| NYT Connections (extended) | 37.3% | — |
| ARC-AGI-1 | 63.7% | — |
| CritPt | 1.1% | — |
| EnigmaEval | 6% | — |
| EBR-Bench | 2.4% | — |
| BIG-Bench Hard | — | 75.1% |
| HellaSwag | — | 95.3% |
| WinoGrande | — | 87.5% |

## Math

- Claude Sonnet 4.5: 32.3 (#216)
- GPT-4: 10.8 (#309)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 77.8% | 1.1% |
| LMArena Math | 1449 | 1269 |
| MATH Level 5 | 97.7% | 23% |
| FrontierMath (Tiers 1-3) | 23.9% | — |
| FrontierMath Tier 4 | 2.4% | — |
| ProofBench | 19% | — |
| Omni-MATH | 55.3% | — |
| FrontierMath (Feb 2025 set) | 15.2% | — |
| FrontierMath Tier 4 (v1) | 4.2% | — |
| GSM8K | — | 92% |

## Knowledge

- Claude Sonnet 4.5: 48.4 (#76)
- GPT-4: 18.4 (#282)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| GPQA Diamond | 82.3% | 35.7% |
| LMArena Expert | 1482 | 1211 |
| Humanity's Last Exam | 13.7% | — |
| SimpleQA Verified | 30.7% | — |
| MMLU-Pro | 86.9% | — |
| Vectara Hallucination Rate | 12% | — |
| GPQA (HELM) | 68.6% | — |
| MMLU | — | 86.4% |
| TriviaQA | — | 84.8% |

## Multimodal

- Claude Sonnet 4.5: 34.8 (#89)
- GPT-4: —

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| VPCT | 39.8% | — |
| LMArena Document | 1450 | — |

## Multilingual

- Claude Sonnet 4.5: 53.4 (#69)
- GPT-4: 40.6 (#215)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| LMArena Non-English | 1425 | 1246 |
| LMArena Chinese | 1459 | 1242 |
| LMArena French | 1458 | 1283 |
| LMArena German | 1427 | 1251 |
| LMArena Japanese | 1390 | 1209 |
| LMArena Korean | 1403 | 1184 |
| LMArena Russian | 1437 | 1251 |
| LMArena Spanish | 1457 | 1261 |

## Instruction Following

- Claude Sonnet 4.5: 75.0 (#78)
- GPT-4: 65.3 (#222)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1459 | 1241 |
| IFEval | 85% | — |

## Long Context

- Claude Sonnet 4.5: 45.2 (#46)
- GPT-4: 37.7 (#212)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1476 | 1244 |

## Writing & Preference

- Claude Sonnet 4.5: 66.5 (#34)
- GPT-4: 34.9 (#268)

| Benchmark | Claude Sonnet 4.5 | GPT-4 |
|---|---|---|
| LMArena Text | 1439 | 1263 |
| LMArena Creative Writing | 1442 | 1244 |
| EQ-Bench Creative Writing | 1678 | 752 |
| LMArena Multi-Turn | 1465 | 1257 |
| WildBench | 85.4% | — |

## FAQ

### Is Claude Sonnet 4.5 better than GPT-4?

Claude Sonnet 4.5 is the stronger model overall, scoring 44.1 to 29.1 on the Noometry Index.

### Which is cheaper, Claude Sonnet 4.5 or GPT-4?

Claude Sonnet 4.5 is cheaper. It lists at $3 per million input tokens and $15 per million output tokens; GPT-4 lists at $30 and $60.

### Is Claude Sonnet 4.5 or GPT-4 better for coding?

Claude Sonnet 4.5 scores higher on coding benchmarks: 47.3 versus 31.6 in the Noometry coding category.

### Which has the bigger context window?

Claude Sonnet 4.5 does, with 200K tokens against 8K.

### How many benchmarks do Claude Sonnet 4.5 and GPT-4 share?

29 benchmarks have published results for both models. Claude Sonnet 4.5 has 73 scored results on Noometry and GPT-4 has 38.