# DeepSeek-V3.1 vs GPT-5.4 mini

> GPT-5.4 mini is the stronger model overall, scoring 44.8 to 42.2 on the Noometry Index. DeepSeek-V3.1 costs 4.0× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/deepseek-v3-1-vs-gpt-5-4-mini
- Last updated: 2026-10-10
- Shared benchmarks: 25

## Summary

- They share 25 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 0 categories and GPT-5.4 mini in 8 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 mini leads 45.4 to 36.1.
- The biggest single-benchmark swing is WeirdML: 38.4% for DeepSeek-V3.1 and 60.3% for GPT-5.4 mini.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $0.75 / $4.50 for GPT-5.4 mini.
- GPT-5.4 mini accepts more context: 400K tokens versus 164K.
- DeepSeek-V3.1 has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.2 | 44.8 |
| Rank | 106 | 69 |
| Context | 164K | 400K |
| Input $/M | $0.25 | $0.75 |
| Output $/M | $0.95 | $4.50 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek-V3.1: 39.8 (#140)
- GPT-5.4 mini: 45.0 (#67)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| WeirdML | 38.4% | 60.3% |
| LMArena Coding | 1417 | 1438 |
| FrontierCode | — | 27% |
| LMArena WebDev | — | 1397 |
| SciCode | — | 49.9% |
| ALE-Bench | — | 1,189 |

## Agentic & Tool Use

- DeepSeek-V3.1: —
- GPT-5.4 mini: 29.6 (#74)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| DeepResearch Bench | — | 36.3% |

## Reasoning

- DeepSeek-V3.1: 27.3 (#109)
- GPT-5.4 mini: 30.3 (#77)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| Kagi LLM Benchmark | 53.2% | 37.9% |
| LMArena Hard Prompts | 1417 | 1424 |
| DTBench | 82.7% | 80% |
| LMCA | 24.3% | 40.8% |
| Epoch Capabilities Index | 139.92 | 148.83 |
| ForecastBench | 58 | 57 |
| ARC-AGI-2 | — | 18.9% |
| SimpleBench | 40% | — |
| NYT Connections (extended) | — | 61.8% |
| ARC-AGI-1 | — | 63.7% |
| CritPt | — | 10% |
| Chess Puzzles | — | 24% |
| Thematic Generalization | — | 61.7% |
| Mystery Game Puzzles | — | 11% |

## Math

- DeepSeek-V3.1: 36.1 (#125)
- GPT-5.4 mini: 45.4 (#63)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| LMArena Math | 1420 | 1418 |
| FrontierMath (Tiers 1-3) | — | 51.2% |
| FrontierMath Tier 4 | — | 9.8% |
| OTIS Mock AIME 2024-2025 | — | 88.9% |
| ProofBench | — | 21% |
| FrontierMath (Feb 2025 set) | — | 28.3% |
| FrontierMath Tier 4 (v1) | — | 2.1% |

## Knowledge

- DeepSeek-V3.1: 43.0 (#86)
- GPT-5.4 mini: 51.4 (#62)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| Vectara Hallucination Rate | 5.5% | 5.5% |
| LMArena Expert | 1403 | 1434 |
| GPQA Diamond | — | 86.9% |
| SimpleQA Verified | — | 29.4% |

## Multimodal

- DeepSeek-V3.1: —
- GPT-5.4 mini: 39.0 (#51)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | — | 1245 |

## Multilingual

- DeepSeek-V3.1: 51.8 (#100)
- GPT-5.4 mini: 52.1 (#94)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1400 | 1405 |
| LMArena Chinese | 1470 | 1446 |
| LMArena French | 1447 | 1440 |
| LMArena German | 1412 | 1407 |
| LMArena Japanese | 1374 | 1375 |
| LMArena Korean | 1337 | 1369 |
| LMArena Russian | 1405 | 1417 |
| LMArena Spanish | 1429 | 1403 |

## Instruction Following

- DeepSeek-V3.1: 74.1 (#105)
- GPT-5.4 mini: 74.3 (#99)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1400 | 1405 |

## Long Context

- DeepSeek-V3.1: 36.3 (#227)
- GPT-5.4 mini: 42.9 (#107)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1422 | 1407 |
| Fiction.LiveBench | 52.8% | — |

## Writing & Preference

- DeepSeek-V3.1: 60.5 (#93)
- GPT-5.4 mini: 64.2 (#56)

| Benchmark | DeepSeek-V3.1 | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1420 | 1411 |
| LMArena Creative Writing | 1401 | 1370 |
| EQ-Bench Creative Writing | 1436 | 1665 |
| LMArena Multi-Turn | 1408 | 1429 |

## FAQ

### Is DeepSeek-V3.1 better than GPT-5.4 mini?

GPT-5.4 mini is the stronger model overall, scoring 44.8 to 42.2 on the Noometry Index. DeepSeek-V3.1 costs 4.0× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.

### Which is cheaper, DeepSeek-V3.1 or GPT-5.4 mini?

DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.

### Is DeepSeek-V3.1 or GPT-5.4 mini better for coding?

GPT-5.4 mini scores higher on coding benchmarks: 45.0 versus 39.8 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.4 mini does, with 400K tokens against 164K.

### How many benchmarks do DeepSeek-V3.1 and GPT-5.4 mini share?

25 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and GPT-5.4 mini has 46.