# DeepSeek V4 Pro vs GPT-5 Mini

> DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 41.8 on the Noometry Index.

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

## Summary

- They share 40 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 8 categories and GPT-5 Mini in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 23.9.
- The biggest single-benchmark swing is ARC-AGI-2: 61.3% for DeepSeek V4 Pro and 4.4% for GPT-5 Mini.
- GPT-5 Mini is cheaper at $0.25 / $2 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
- 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 Mini |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 54.3 | 41.8 |
| Rank | 31 | 128 |
| Context | 1M | 400K |
| Input $/M | $0.66 | $0.25 |
| Output $/M | $1.98 | $2 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek V4 Pro: 52.4 (#34)
- GPT-5 Mini: 40.1 (#146)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| SWE-bench Verified | 77.6% | 64.7% |
| SciCode | 51% | 39.2% |
| WeirdML | 66.2% | 52.7% |
| LMArena Coding | 1470 | 1406 |
| ALE-Bench | 1,403 | 799.77 |
| FrontierCode | 28.6% | — |
| SWE-bench Verified (bash only) | — | 59.8% |
| LMArena WebDev | 1582 | — |
| SWE-bench Multilingual | — | 39.7% |
| AlgoTune | — | 1.38 |

## Agentic & Tool Use

- DeepSeek V4 Pro: 32.8 (#58)
- GPT-5 Mini: 31.1 (#70)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| Vending-Bench 2 | 3,285 | -31.18 |
| Terminal-Bench | — | 34.8% |
| APEX-Agents | 47.3% | — |
| Berkeley Function Calling Leaderboard | — | 55.5% |

## Reasoning

- DeepSeek V4 Pro: 56.5 (#24)
- GPT-5 Mini: 23.9 (#168)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| ARC-AGI-2 | 61.3% | 4.4% |
| Kagi LLM Benchmark | 53.5% | 70.3% |
| ARC-AGI-1 | 90.5% | 54.3% |
| CritPt | 18% | 0% |
| Chess Puzzles | 47% | 30% |
| LMArena Hard Prompts | 1461 | 1380 |
| Mystery Game Puzzles | 43% | 10% |
| DTBench | 93.9% | 80.5% |
| LMCA | 45.5% | 34.2% |
| Epoch Capabilities Index | 155.31 | 145.52 |
| ForecastBench | 56.1 | 61 |
| NYT Connections (extended) | 91.3% | — |
| EnigmaEval | — | 8.2% |
| Surface Evolver Bench | 40% | — |

## Math

- DeepSeek V4 Pro: 64.8 (#30)
- GPT-5 Mini: 46.7 (#69)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 64.6% | 46.7% |
| FrontierMath Tier 4 | 26.8% | 12.2% |
| OTIS Mock AIME 2024-2025 | 98.6% | 86.7% |
| ProofBench | 50% | 9% |
| LMArena Math | 1455 | 1378 |
| MathArena Final-Answer Competitions | 76.6% | — |
| Omni-MATH | — | 72.2% |
| MATH Level 5 | — | 97.8% |
| FrontierMath (Feb 2025 set) | — | 27.2% |
| FrontierMath Tier 4 (v1) | — | 6.3% |

## Knowledge

- DeepSeek V4 Pro: 59.5 (#31)
- GPT-5 Mini: 45.6 (#86)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| GPQA Diamond | 91.7% | 75% |
| SimpleQA Verified | 52.9% | 21.6% |
| Vectara Hallucination Rate | 8.6% | 12.9% |
| LMArena Expert | 1464 | 1379 |
| Humanity's Last Exam | — | 19.4% |
| MMLU-Pro | — | 83.5% |
| Confabulations | — | 13.3% |
| GPQA (HELM) | — | 75.6% |

## Multimodal

- DeepSeek V4 Pro: —
- GPT-5 Mini: 35.6 (#85)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| LMArena Vision | — | 1202 |
| VPCT | — | 40.2% |

## Multilingual

- DeepSeek V4 Pro: 54.4 (#45)
- GPT-5 Mini: 48.9 (#137)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| LMArena Non-English | 1439 | 1363 |
| LMArena Chinese | 1486 | 1385 |
| LMArena French | 1472 | 1386 |
| LMArena German | 1458 | 1366 |
| LMArena Japanese | 1445 | 1341 |
| LMArena Korean | 1447 | 1308 |
| LMArena Russian | 1453 | 1362 |
| LMArena Spanish | 1458 | 1355 |

## Instruction Following

- DeepSeek V4 Pro: 76.1 (#47)
- GPT-5 Mini: 76.2 (#46)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| LMArena Instruction Following | 1448 | 1357 |
| IFEval | — | 92.7% |

## Long Context

- DeepSeek V4 Pro: 45.0 (#51)
- GPT-5 Mini: 41.9 (#132)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| LMArena Longer Query | 1458 | 1355 |
| Fiction.LiveBench | — | 69.4% |
| CL-bench Life | 13.5% | — |

## Writing & Preference

- DeepSeek V4 Pro: 65.5 (#46)
- GPT-5 Mini: 55.2 (#148)

| Benchmark | DeepSeek V4 Pro | GPT-5 Mini |
|---|---|---|
| LMArena Text | 1451 | 1373 |
| LMArena Creative Writing | 1446 | 1325 |
| EQ-Bench Creative Writing | 1553 | 1313 |
| LMArena Multi-Turn | 1467 | 1363 |
| Short-Story Creative Writing | — | 83.1% |
| WildBench | — | 85.5% |
| EQ-Bench 4 | 1166 | — |

## FAQ

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

DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 41.8 on the Noometry Index.

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

GPT-5 Mini is cheaper. It lists at $0.25 per million input tokens and $2 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.

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

DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 40.1 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 Mini share?

40 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and GPT-5 Mini has 60.