# DeepSeek V4 Flash vs GPT-4

> DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 28.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-flash-vs-gpt-4
- Last updated: 2026-10-10
- Shared benchmarks: 26

## Summary

- They share 26 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 8 categories and GPT-4 in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.0 to 10.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 94.4% for DeepSeek V4 Flash and 1.1% for GPT-4.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $30 / $60 for GPT-4.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 8K.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 53.6 | 28.9 |
| Rank | 32 | 308 |
| Context | 1M | 8K |
| Input $/M | $0.15 | $30 |
| Output $/M | $0.60 | $60 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek V4 Flash: 47.6 (#56)
- GPT-4: 31.2 (#277)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| WeirdML | 63% | 12.4% |
| LMArena Coding | 1457 | 1254 |
| FrontierCode | 18.8% | — |
| LMArena WebDev | 1581 | — |
| SciCode | 49.9% | — |
| BigCodeBench Instruct | — | 46% |
| BigCodeBench Complete | — | 57.2% |
| ALE-Bench | 1,306 | — |
| HumanEval+ | — | 79.3% |

## Agentic & Tool Use

- DeepSeek V4 Flash: —
- GPT-4: —

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| METR Time Horizons | — | 36.1% |

## Reasoning

- DeepSeek V4 Flash: 53.7 (#25)
- GPT-4: 17.3 (#285)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| Chess Puzzles | 33% | 4% |
| LMArena Hard Prompts | 1443 | 1241 |
| Mystery Game Puzzles | 34% | 12% |
| DTBench | 90.9% | 62.7% |
| LMCA | 41.7% | 17.1% |
| Epoch Capabilities Index | 154.51 | 125.88 |
| ARC-AGI-2 | 61.4% | — |
| SimpleBench | 61.1% | — |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 89.6% | — |
| ARC-AGI-1 | 89% | — |
| CritPt | 16.6% | — |
| BIG-Bench Hard | — | 75.1% |
| ForecastBench | — | 57.8 |
| HellaSwag | — | 95.3% |
| WinoGrande | — | 87.5% |

## Math

- DeepSeek V4 Flash: 60.0 (#31)
- GPT-4: 10.1 (#304)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 94.4% | 1.1% |
| LMArena Math | 1427 | 1269 |
| FrontierMath (Tiers 1-3) | 57.5% | — |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
| ProofBench | 56% | — |
| MATH Level 5 | — | 23% |
| GSM8K | — | 92% |

## Knowledge

- DeepSeek V4 Flash: 55.5 (#43)
- GPT-4: 18.4 (#277)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| GPQA Diamond | 91% | 35.7% |
| LMArena Expert | 1441 | 1211 |
| SimpleQA Verified | 33.6% | — |
| MMLU | — | 86.4% |
| TriviaQA | — | 84.8% |

## Multilingual

- DeepSeek V4 Flash: 53.2 (#71)
- GPT-4: 40.8 (#210)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| LMArena Non-English | 1420 | 1246 |
| LMArena Chinese | 1467 | 1242 |
| LMArena French | 1438 | 1282 |
| LMArena German | 1418 | 1251 |
| LMArena Japanese | 1403 | 1209 |
| LMArena Korean | 1383 | 1184 |
| LMArena Russian | 1428 | 1251 |
| LMArena Spanish | 1437 | 1261 |

## Instruction Following

- DeepSeek V4 Flash: 75.1 (#76)
- GPT-4: 65.5 (#217)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1421 | 1241 |

## Long Context

- DeepSeek V4 Flash: 43.8 (#81)
- GPT-4: 37.7 (#207)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1435 | 1244 |

## Writing & Preference

- DeepSeek V4 Flash: 64.0 (#59)
- GPT-4: 35.0 (#263)

| Benchmark | DeepSeek V4 Flash | GPT-4 |
|---|---|---|
| LMArena Text | 1432 | 1263 |
| LMArena Creative Writing | 1403 | 1244 |
| EQ-Bench Creative Writing | 1559 | 752 |
| LMArena Multi-Turn | 1449 | 1257 |

## FAQ

### Is DeepSeek V4 Flash better than GPT-4?

DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 28.9 on the Noometry Index.

### Which is cheaper, DeepSeek V4 Flash or GPT-4?

DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; GPT-4 lists at $30 and $60.

### Is DeepSeek V4 Flash or GPT-4 better for coding?

DeepSeek V4 Flash scores higher on coding benchmarks: 47.6 versus 31.2 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Flash does, with 1M tokens against 8K.

### How many benchmarks do DeepSeek V4 Flash and GPT-4 share?

26 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and GPT-4 has 38.