# DeepSeek-V3.1 vs GPT-4.5

> DeepSeek-V3.1 is the stronger model overall, scoring 42.2 to 36.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v3-1-vs-gpt-4-5
- Last updated: 2026-10-10
- Shared benchmarks: 22

## Summary

- They share 22 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 5 categories and GPT-4.5 in 3 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek-V3.1 leads 27.3 to 14.1.
- The biggest single-benchmark swing is Fiction.LiveBench: 52.8% for DeepSeek-V3.1 and 63.9% for GPT-4.5.
- DeepSeek-V3.1 has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.2 | 36.9 |
| Rank | 106 | 200 |
| Context | 164K | — |
| Input $/M | $0.25 | — |
| Output $/M | $0.95 | — |
| Weights | Open | Proprietary |

## Coding

- DeepSeek-V3.1: 39.8 (#140)
- GPT-4.5: 41.8 (#102)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| WeirdML | 38.4% | 39.4% |
| LMArena Coding | 1417 | 1397 |
| Aider Polyglot | — | 44.9% |
| LiveBench Coding | — | 75.2% |

## Agentic & Tool Use

- DeepSeek-V3.1: —
- GPT-4.5: 27.6 (#92)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| Cybench | — | 17.5% |

## Reasoning

- DeepSeek-V3.1: 27.3 (#109)
- GPT-4.5: 14.1 (#320)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| SimpleBench | 40% | 34.5% |
| LMArena Hard Prompts | 1417 | 1404 |
| Epoch Capabilities Index | 139.92 | 136.74 |
| ForecastBench | 58 | 61.7 |
| ARC-AGI-2 | — | 0.8% |
| Kagi LLM Benchmark | 53.2% | — |
| ARC-AGI-1 | — | 10.3% |
| EnigmaEval | — | 3.2% |
| LiveBench Reasoning | — | 71.1% |
| DTBench | 82.7% | — |
| LiveBench Data Analysis | — | 64.3% |
| LMCA | 24.3% | — |
| LiveBench | — | 69% |

## Math

- DeepSeek-V3.1: 36.1 (#125)
- GPT-4.5: 31.0 (#200)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| LMArena Math | 1420 | 1412 |
| OTIS Mock AIME 2024-2025 | — | 37.8% |
| LiveBench Math | — | 69.3% |
| MATH Level 5 | — | 78.6% |

## Knowledge

- DeepSeek-V3.1: 43.0 (#86)
- GPT-4.5: 32.4 (#205)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| LMArena Expert | 1403 | 1394 |
| GPQA Diamond | — | 68.7% |
| Humanity's Last Exam | — | 5.4% |
| Confabulations | — | 13.6% |
| Vectara Hallucination Rate | 5.5% | — |

## Multimodal

- DeepSeek-V3.1: —
- GPT-4.5: 37.0 (#68)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| LMArena Vision | — | 1195 |
| VPCT | — | 45% |

## Multilingual

- DeepSeek-V3.1: 51.8 (#100)
- GPT-4.5: 52.7 (#78)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| LMArena Non-English | 1400 | 1413 |
| LMArena Chinese | 1470 | 1421 |
| LMArena French | 1447 | 1417 |
| LMArena German | 1412 | 1457 |
| LMArena Japanese | 1374 | 1416 |
| LMArena Korean | 1337 | 1392 |
| LMArena Russian | 1405 | 1419 |
| LMArena Spanish | 1429 | — |

## Instruction Following

- DeepSeek-V3.1: 74.1 (#105)
- GPT-4.5: 72.8 (#129)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1404 |
| LiveBench Instruction Following | — | 72.3% |

## Long Context

- DeepSeek-V3.1: 36.3 (#227)
- GPT-4.5: 40.4 (#150)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| Fiction.LiveBench | 52.8% | 63.9% |
| LMArena Longer Query | 1422 | 1406 |

## Writing & Preference

- DeepSeek-V3.1: 60.5 (#93)
- GPT-4.5: 57.1 (#129)

| Benchmark | DeepSeek-V3.1 | GPT-4.5 |
|---|---|---|
| LMArena Text | 1420 | 1417 |
| LMArena Creative Writing | 1401 | 1394 |
| EQ-Bench Creative Writing | 1436 | 1258 |
| LMArena Multi-Turn | 1408 | 1444 |
| Short-Story Creative Writing | — | 75.6% |
| LiveBench Language | — | 61.5% |

## FAQ

### Is DeepSeek-V3.1 better than GPT-4.5?

DeepSeek-V3.1 is the stronger model overall, scoring 42.2 to 36.9 on the Noometry Index.

### Is DeepSeek-V3.1 or GPT-4.5 better for coding?

GPT-4.5 scores higher on coding benchmarks: 41.8 versus 39.8 in the Noometry coding category.

### How many benchmarks do DeepSeek-V3.1 and GPT-4.5 share?

22 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and GPT-4.5 has 42.