# DeepSeek V4 Pro vs GPT-5

> DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 50.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-pro-vs-gpt-5
- Last updated: 2026-10-11
- Shared benchmarks: 40

## Summary

- They share 40 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 7 categories and GPT-5 in 2 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-5 leads 69.5 to 45.0.
- The biggest single-benchmark swing is ARC-AGI-2: 61.3% for DeepSeek V4 Pro and 9.9% for GPT-5.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $1.25 / $10 for GPT-5.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 400K.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 54.3 | 50.9 |
| Rank | 31 | 45 |
| Context | 1M | 400K |
| Input $/M | $0.66 | $1.25 |
| Output $/M | $1.98 | $10 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek V4 Pro: 52.4 (#34)
- GPT-5: 50.3 (#47)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| SWE-bench Verified | 77.6% | 73.6% |
| LMArena WebDev | 1582 | 1418 |
| SciCode | 51% | 42.9% |
| WeirdML | 66.2% | 60.7% |
| LMArena Coding | 1470 | 1436 |
| ALE-Bench | 1,403 | 1,162 |
| FrontierCode | 28.6% | — |
| SWE-bench Verified (bash only) | — | 65% |
| Aider Polyglot | — | 88% |
| GSO | — | 6.9% |
| AlgoTune | — | 1.67 |

## Agentic & Tool Use

- DeepSeek V4 Pro: 32.8 (#58)
- GPT-5: 33.1 (#56)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| Terminal-Bench | — | 49.6% |
| APEX-Agents | 47.3% | — |
| GDPval | — | 34.8% |
| Remote Labor Index | — | 1.7% |
| DeepResearch Bench | — | 49.6% |
| BALROG | — | 32.8% |
| LMArena Search | — | 1133 |
| METR Time Horizons | — | 69.6% |
| Vending-Bench 2 | 3,285 | — |

## Reasoning

- DeepSeek V4 Pro: 56.5 (#24)
- GPT-5: 38.3 (#64)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| ARC-AGI-2 | 61.3% | 9.9% |
| Kagi LLM Benchmark | 53.5% | 72.7% |
| ARC-AGI-1 | 90.5% | 65.7% |
| CritPt | 18% | 12.6% |
| Chess Puzzles | 47% | 37% |
| LMArena Hard Prompts | 1461 | 1416 |
| Mystery Game Puzzles | 43% | 23% |
| DTBench | 93.9% | 90.7% |
| LMCA | 45.5% | 40% |
| Epoch Capabilities Index | 155.31 | 150 |
| ForecastBench | 56.1 | 61.4 |
| SimpleBench | — | 56.7% |
| NYT Connections (extended) | 91.3% | — |
| EnigmaEval | — | 10.5% |
| EBR-Bench | — | 12.7% |
| Surface Evolver Bench | 40% | — |

## Math

- DeepSeek V4 Pro: 64.8 (#30)
- GPT-5: 55.0 (#44)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 64.6% | 55.4% |
| FrontierMath Tier 4 | 26.8% | 22% |
| OTIS Mock AIME 2024-2025 | 98.6% | 91.4% |
| ProofBench | 50% | 18% |
| LMArena Math | 1455 | 1407 |
| MathArena Final-Answer Competitions | 76.6% | — |
| Omni-MATH | — | 64.7% |
| MATH Level 5 | — | 98.1% |
| FrontierMath (Feb 2025 set) | — | 32.4% |
| FrontierMath Tier 4 (v1) | — | 12.5% |

## Knowledge

- DeepSeek V4 Pro: 59.5 (#31)
- GPT-5: 56.6 (#43)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| GPQA Diamond | 91.7% | 86.2% |
| SimpleQA Verified | 52.9% | 50.1% |
| Vectara Hallucination Rate | 8.6% | 14.7% |
| LMArena Expert | 1464 | 1419 |
| Humanity's Last Exam | — | 25.3% |
| MMLU-Pro | — | 86.3% |
| Confabulations | — | 10.3% |
| GPQA (HELM) | — | 79.2% |

## Multimodal

- DeepSeek V4 Pro: —
- GPT-5: 46.8 (#13)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| LMArena Vision | — | 1232 |
| GeoBench | — | 81% |
| VPCT | — | 66% |

## Multilingual

- DeepSeek V4 Pro: 54.4 (#45)
- GPT-5: 51.4 (#110)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| LMArena Non-English | 1439 | 1397 |
| LMArena Chinese | 1486 | 1422 |
| LMArena French | 1472 | 1410 |
| LMArena German | 1458 | 1416 |
| LMArena Japanese | 1445 | 1409 |
| LMArena Korean | 1447 | 1360 |
| LMArena Russian | 1453 | 1406 |
| LMArena Spanish | 1458 | 1399 |

## Instruction Following

- DeepSeek V4 Pro: 76.1 (#47)
- GPT-5: 73.8 (#113)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1388 |
| IFEval | — | 87.5% |

## Long Context

- DeepSeek V4 Pro: 45.0 (#51)
- GPT-5: 69.5 (#2)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| LMArena Longer Query | 1458 | 1399 |
| Fiction.LiveBench | — | 97.2% |
| CL-bench Life | 13.5% | — |

## Writing & Preference

- DeepSeek V4 Pro: 65.5 (#46)
- GPT-5: 63.4 (#65)

| Benchmark | DeepSeek V4 Pro | GPT-5 |
|---|---|---|
| LMArena Text | 1451 | 1406 |
| LMArena Creative Writing | 1446 | 1365 |
| EQ-Bench Creative Writing | 1553 | 1627 |
| LMArena Multi-Turn | 1467 | 1426 |
| Short-Story Creative Writing | — | 86% |
| WildBench | — | 85.7% |
| EQ-Bench 4 | 1166 | — |

## FAQ

### Is DeepSeek V4 Pro better than GPT-5?

DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 50.9 on the Noometry Index.

### Which is cheaper, DeepSeek V4 Pro or GPT-5?

DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; GPT-5 lists at $1.25 and $10.

### Is DeepSeek V4 Pro or GPT-5 better for coding?

DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 50.3 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Pro does, with 1M tokens against 400K.

### How many benchmarks do DeepSeek V4 Pro and GPT-5 share?

40 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and GPT-5 has 69.