# DeepSeek-V3.1 vs GPT-4

> DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 29.1 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v3-1-vs-gpt-4
- Last updated: 2026-10-11
- Shared benchmarks: 23

## Summary

- They share 23 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 7 categories and GPT-4 in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3.1 leads 38.9 to 10.8.
- The biggest single-benchmark swing is WeirdML: 38.4% for DeepSeek-V3.1 and 12.4% for GPT-4.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $30 / $60 for GPT-4.
- DeepSeek-V3.1 accepts more context: 164K tokens versus 8K.
- DeepSeek-V3.1 has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.8 | 29.1 |
| Rank | 108 | 316 |
| Context | 164K | 8K |
| Input $/M | $0.25 | $30 |
| Output $/M | $0.95 | $60 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek-V3.1: 40.3 (#144)
- GPT-4: 31.6 (#283)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| WeirdML | 38.4% | 12.4% |
| LMArena Coding | 1417 | 1254 |
| BigCodeBench Instruct | — | 46% |
| BigCodeBench Complete | — | 57.2% |
| HumanEval+ | — | 79.3% |

## Agentic & Tool Use

- DeepSeek-V3.1: —
- GPT-4: —

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| METR Time Horizons | — | 36.1% |

## Reasoning

- DeepSeek-V3.1: 27.9 (#110)
- GPT-4: 17.8 (#289)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| LMArena Hard Prompts | 1417 | 1241 |
| DTBench | 82.7% | 62.7% |
| LMCA | 24.3% | 17.1% |
| Epoch Capabilities Index | 139.92 | 125.89 |
| ForecastBench | 58 | 57.8 |
| SimpleBench | 40% | — |
| Kagi LLM Benchmark | 53.2% | — |
| Chess Puzzles | — | 4% |
| Mystery Game Puzzles | — | 12% |
| BIG-Bench Hard | — | 75.1% |
| HellaSwag | — | 95.3% |
| WinoGrande | — | 87.5% |

## Math

- DeepSeek-V3.1: 38.9 (#122)
- GPT-4: 10.8 (#309)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| LMArena Math | 1420 | 1269 |
| OTIS Mock AIME 2024-2025 | — | 1.1% |
| MATH Level 5 | — | 23% |
| GSM8K | — | 92% |

## Knowledge

- DeepSeek-V3.1: 43.7 (#90)
- GPT-4: 18.4 (#282)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| LMArena Expert | 1405 | 1211 |
| GPQA Diamond | — | 35.7% |
| Vectara Hallucination Rate | 5.5% | — |
| MMLU | — | 86.4% |
| TriviaQA | — | 84.8% |

## Multilingual

- DeepSeek-V3.1: 51.6 (#106)
- GPT-4: 40.6 (#215)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| LMArena Non-English | 1400 | 1246 |
| LMArena Chinese | 1469 | 1242 |
| LMArena French | 1447 | 1283 |
| LMArena German | 1411 | 1251 |
| LMArena Japanese | 1378 | 1209 |
| LMArena Korean | 1337 | 1184 |
| LMArena Russian | 1405 | 1251 |
| LMArena Spanish | 1431 | 1261 |

## Instruction Following

- DeepSeek-V3.1: 73.9 (#110)
- GPT-4: 65.3 (#222)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1241 |

## Long Context

- DeepSeek-V3.1: 36.3 (#232)
- GPT-4: 37.7 (#212)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1422 | 1244 |
| Fiction.LiveBench | 52.8% | — |

## Writing & Preference

- DeepSeek-V3.1: 60.3 (#98)
- GPT-4: 34.9 (#268)

| Benchmark | DeepSeek-V3.1 | GPT-4 |
|---|---|---|
| LMArena Text | 1420 | 1263 |
| LMArena Creative Writing | 1401 | 1244 |
| EQ-Bench Creative Writing | 1436 | 752 |
| LMArena Multi-Turn | 1408 | 1257 |

## FAQ

### Is DeepSeek-V3.1 better than GPT-4?

DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 29.1 on the Noometry Index.

### Which is cheaper, DeepSeek-V3.1 or GPT-4?

DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; GPT-4 lists at $30 and $60.

### Is DeepSeek-V3.1 or GPT-4 better for coding?

DeepSeek-V3.1 scores higher on coding benchmarks: 40.3 versus 31.6 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek-V3.1 does, with 164K tokens against 8K.

### How many benchmarks do DeepSeek-V3.1 and GPT-4 share?

23 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and GPT-4 has 38.