# DeepSeek V4 Flash vs GPT-4o

> DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 28.5 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-flash-vs-gpt-4o
- Last updated: 2026-10-10
- Shared benchmarks: 30

## Summary

- They share 30 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 8 categories and GPT-4o in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.0 to 10.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 94.4% for DeepSeek V4 Flash and 6.4% for GPT-4o.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $2.50 / $10 for GPT-4o.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 128K.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 53.6 | 28.5 |
| Rank | 32 | 316 |
| Context | 1M | 128K |
| Input $/M | $0.15 | $2.50 |
| Output $/M | $0.60 | $10 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek V4 Flash: 47.6 (#56)
- GPT-4o: 25.1 (#323)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| WeirdML | 63% | 25.1% |
| LMArena Coding | 1457 | 1297 |
| SWE-bench Verified | — | 31% |
| FrontierCode | 18.8% | — |
| SWE-bench Verified (bash only) | — | 21.6% |
| Aider Polyglot | — | 45.3% |
| LMArena WebDev | 1581 | — |
| SciCode | 49.9% | — |
| GSO | — | 0% |
| BigCodeBench Instruct | — | 51.1% |
| LiveBench Coding | — | 51.4% |
| BigCodeBench Complete | — | 61.1% |
| CadEval | — | 26% |
| ALE-Bench | 1,306 | — |
| HumanEval+ | — | 87.2% |
| MBPP+ | — | 72.2% |

## Agentic & Tool Use

- DeepSeek V4 Flash: —
- GPT-4o: 20.8 (#133)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| GDPval | — | 9.9% |
| TheAgentCompany | — | 8.6% |
| Cybench | — | 12.5% |
| BALROG | — | 32.3% |
| LMArena Search | — | 1006 |
| METR Time Horizons | — | 40.8% |

## Reasoning

- DeepSeek V4 Flash: 53.7 (#25)
- GPT-4o: 9.4 (#338)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| ARC-AGI-2 | 61.4% | 0% |
| SimpleBench | 61.1% | 17.8% |
| ARC-AGI-1 | 89% | 4.5% |
| CritPt | 16.6% | 0% |
| Chess Puzzles | 33% | 13% |
| LMArena Hard Prompts | 1443 | 1281 |
| DTBench | 90.9% | 64.5% |
| LMCA | 41.7% | 16.6% |
| Epoch Capabilities Index | 154.51 | 128.97 |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 89.6% | — |
| EnigmaEval | — | 0.8% |
| LiveBench Reasoning | — | 55.8% |
| Mystery Game Puzzles | 34% | — |
| LiveBench Data Analysis | — | 60.9% |
| ForecastBench | — | 57.7 |
| LiveBench | — | 55.3% |

## Math

- DeepSeek V4 Flash: 60.0 (#31)
- GPT-4o: 10.2 (#303)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| FrontierMath (Tiers 1-3) | 57.5% | 0.4% |
| OTIS Mock AIME 2024-2025 | 94.4% | 6.4% |
| LMArena Math | 1427 | 1285 |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
| ProofBench | 56% | — |
| Omni-MATH | — | 29.3% |
| LiveBench Math | — | 49.5% |
| MATH Level 5 | — | 53.3% |
| FrontierMath (Feb 2025 set) | — | 0.3% |

## Knowledge

- DeepSeek V4 Flash: 55.5 (#43)
- GPT-4o: 28.8 (#235)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| GPQA Diamond | 91% | 49.2% |
| SimpleQA Verified | 33.6% | 26% |
| LMArena Expert | 1441 | 1250 |
| Humanity's Last Exam | — | 2.7% |
| MMLU-Pro | — | 71.3% |
| Confabulations | — | 15.3% |
| Vectara Hallucination Rate | — | 9.6% |
| GPQA (HELM) | — | 52% |
| MMLU | — | 88.1% |

## Multimodal

- DeepSeek V4 Flash: —
- GPT-4o: 33.9 (#85)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| LMArena Vision | — | 1137 |
| Video-MME | — | 71.9% |
| GeoBench | — | 71% |
| VPCT | — | 40% |
| ScienceQA | — | 88.5% |

## Multilingual

- DeepSeek V4 Flash: 53.2 (#71)
- GPT-4o: 43.4 (#181)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| LMArena Non-English | 1420 | 1283 |
| LMArena Chinese | 1467 | 1277 |
| LMArena French | 1438 | 1303 |
| LMArena German | 1418 | 1282 |
| LMArena Japanese | 1403 | 1257 |
| LMArena Korean | 1383 | 1234 |
| LMArena Russian | 1428 | 1286 |
| LMArena Spanish | 1437 | 1292 |

## Instruction Following

- DeepSeek V4 Flash: 75.1 (#76)
- GPT-4o: 66.8 (#202)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| LMArena Instruction Following | 1421 | 1278 |
| LiveBench Instruction Following | — | 68.6% |
| IFEval | — | 81.7% |

## Long Context

- DeepSeek V4 Flash: 43.8 (#81)
- GPT-4o: 39.4 (#174)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| LMArena Longer Query | 1435 | 1289 |
| Fiction.LiveBench | — | 66.7% |

## Writing & Preference

- DeepSeek V4 Flash: 64.0 (#59)
- GPT-4o: 52.8 (#160)

| Benchmark | DeepSeek V4 Flash | GPT-4o |
|---|---|---|
| LMArena Text | 1432 | 1300 |
| LMArena Creative Writing | 1403 | 1291 |
| LMArena Multi-Turn | 1449 | 1302 |
| Short-Story Creative Writing | — | 81.8% |
| EQ-Bench Creative Writing | 1559 | — |
| WildBench | — | 82.8% |
| LiveBench Language | — | 47.6% |

## FAQ

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

DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 28.5 on the Noometry Index.

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

DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; GPT-4o lists at $2.50 and $10.

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

DeepSeek V4 Flash scores higher on coding benchmarks: 47.6 versus 25.1 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Flash does, with 1M tokens against 128K.

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

30 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and GPT-4o has 72.