# DeepSeek V4 Pro vs GPT-5.4 mini

> DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 45.0 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-pro-vs-gpt-5-4-mini
- Last updated: 2026-10-10
- Shared benchmarks: 41

## Summary

- They share 41 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and GPT-5.4 mini in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 30.4.
- The biggest single-benchmark swing is ARC-AGI-2: 61.3% for DeepSeek V4 Pro and 18.9% for GPT-5.4 mini.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $0.75 / $4.50 for GPT-5.4 mini.
- 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-5.4 mini |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 54.3 | 45.0 |
| Rank | 31 | 76 |
| Context | 1M | 400K |
| Input $/M | $0.66 | $0.75 |
| Output $/M | $1.98 | $4.50 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek V4 Pro: 52.4 (#34)
- GPT-5.4 mini: 45.2 (#72)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| FrontierCode | 28.6% | 27% |
| LMArena WebDev | 1582 | 1397 |
| SciCode | 51% | 49.9% |
| WeirdML | 66.2% | 60.3% |
| LMArena Coding | 1470 | 1438 |
| ALE-Bench | 1,403 | 1,189 |
| SWE-bench Verified | 77.6% | — |

## Agentic & Tool Use

- DeepSeek V4 Pro: 32.8 (#58)
- GPT-5.4 mini: 29.9 (#81)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| APEX-Agents | 47.3% | — |
| DeepResearch Bench | — | 36.3% |
| Vending-Bench 2 | 3,285 | — |

## Reasoning

- DeepSeek V4 Pro: 56.5 (#24)
- GPT-5.4 mini: 30.4 (#85)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| ARC-AGI-2 | 61.3% | 18.9% |
| Kagi LLM Benchmark | 53.5% | 37.9% |
| NYT Connections (extended) | 91.3% | 61.8% |
| ARC-AGI-1 | 90.5% | 63.7% |
| CritPt | 18% | 10% |
| Chess Puzzles | 47% | 24% |
| LMArena Hard Prompts | 1461 | 1424 |
| Mystery Game Puzzles | 43% | 11% |
| DTBench | 93.9% | 80% |
| LMCA | 45.5% | 40.8% |
| Epoch Capabilities Index | 155.31 | 148.84 |
| ForecastBench | 56.1 | 57 |
| Thematic Generalization | — | 61.7% |
| Surface Evolver Bench | 40% | — |

## Math

- DeepSeek V4 Pro: 64.8 (#30)
- GPT-5.4 mini: 45.5 (#75)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 64.6% | 51.2% |
| FrontierMath Tier 4 | 26.8% | 9.8% |
| OTIS Mock AIME 2024-2025 | 98.6% | 88.9% |
| ProofBench | 50% | 21% |
| LMArena Math | 1455 | 1419 |
| MathArena Final-Answer Competitions | 76.6% | — |
| FrontierMath (Feb 2025 set) | — | 28.3% |
| FrontierMath Tier 4 (v1) | — | 2.1% |

## Knowledge

- DeepSeek V4 Pro: 59.5 (#31)
- GPT-5.4 mini: 51.5 (#67)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| GPQA Diamond | 91.7% | 86.9% |
| SimpleQA Verified | 52.9% | 29.4% |
| Vectara Hallucination Rate | 8.6% | 5.5% |
| LMArena Expert | 1464 | 1435 |

## Multimodal

- DeepSeek V4 Pro: —
- GPT-5.4 mini: 39.7 (#56)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | — | 1245 |

## Multilingual

- DeepSeek V4 Pro: 54.4 (#45)
- GPT-5.4 mini: 51.9 (#96)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1439 | 1405 |
| LMArena Chinese | 1486 | 1446 |
| LMArena French | 1472 | 1440 |
| LMArena German | 1458 | 1409 |
| LMArena Japanese | 1445 | 1374 |
| LMArena Korean | 1447 | 1368 |
| LMArena Russian | 1453 | 1417 |
| LMArena Spanish | 1458 | 1405 |

## Instruction Following

- DeepSeek V4 Pro: 76.1 (#47)
- GPT-5.4 mini: 74.1 (#102)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1448 | 1405 |

## Long Context

- DeepSeek V4 Pro: 45.0 (#51)
- GPT-5.4 mini: 43.0 (#112)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1458 | 1407 |
| CL-bench Life | 13.5% | — |

## Writing & Preference

- DeepSeek V4 Pro: 65.5 (#46)
- GPT-5.4 mini: 64.0 (#58)

| Benchmark | DeepSeek V4 Pro | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1451 | 1412 |
| LMArena Creative Writing | 1446 | 1370 |
| EQ-Bench Creative Writing | 1553 | 1665 |
| LMArena Multi-Turn | 1467 | 1429 |
| EQ-Bench 4 | 1166 | — |

## FAQ

### Is DeepSeek V4 Pro better than GPT-5.4 mini?

DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 45.0 on the Noometry Index.

### Which is cheaper, DeepSeek V4 Pro or GPT-5.4 mini?

DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.

### Is DeepSeek V4 Pro or GPT-5.4 mini better for coding?

DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 45.2 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-5.4 mini share?

41 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and GPT-5.4 mini has 46.