# DeepSeek-V3 vs GPT-5.6 Luna

> GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 39.2 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v3-vs-gpt-5-6-luna
- Last updated: 2026-10-10
- Shared benchmarks: 28

## Summary

- They share 28 benchmarks with published results for both. DeepSeek-V3 scores higher in 0 categories and GPT-5.6 Luna in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.5 to 30.4.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 98.3% for GPT-5.6 Luna.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $0.20 / $1.20 for GPT-5.6 Luna.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 164K.
- DeepSeek-V3 has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 39.2 | 54.7 |
| Rank | 157 | 24 |
| Context | 164K | 1.05M |
| Input $/M | $0.24 | $0.20 |
| Output $/M | $0.90 | $1.20 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek-V3: 41.9 (#99)
- GPT-5.6 Luna: 53.9 (#25)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| SciCode | 35.8% | 53.6% |
| WeirdML | 36.1% | 60.9% |
| LMArena Coding | 1368 | 1466 |
| DeepSWE | — | 67.2% |
| FrontierCode | — | 39.8% |
| Aider Polyglot | 55.1% | — |
| CursorBench | — | 35.9% |
| LMArena WebDev | — | 1517 |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 1,667 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |

## Agentic & Tool Use

- DeepSeek-V3: —
- GPT-5.6 Luna: 35.3 (#37)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| BALROG | — | 45.6% |
| GDP.pdf | — | 22.7% |
| METR Time Horizons | 49.6% | — |
| Vending-Bench 2 | — | 4,095 |

## Reasoning

- DeepSeek-V3: 20.0 (#226)
- GPT-5.6 Luna: 47.5 (#38)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| SimpleBench | 27.2% | 46.8% |
| Kagi LLM Benchmark | 52.3% | 49.1% |
| CritPt | 0% | 20.6% |
| LMArena Hard Prompts | 1365 | 1452 |
| DTBench | 64.8% | 89.1% |
| LMCA | 15.5% | 48.5% |
| Epoch Capabilities Index | 135.95 | 156.32 |
| ARC-AGI-2 | — | 59.5% |
| NYT Connections (extended) | — | 69.4% |
| ARC-AGI-1 | — | 88% |
| Chess Puzzles | — | 40% |
| LiveBench Reasoning | 65.8% | — |
| Mystery Game Puzzles | — | 21% |
| LiveBench Data Analysis | 60.9% | — |
| Surface Evolver Bench | — | 61.9% |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |

## Math

- DeepSeek-V3: 30.4 (#208)
- GPT-5.6 Luna: 77.5 (#12)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 98.3% |
| LMArena Math | 1373 | 1455 |
| FrontierMath (Tiers 1-3) | — | 82.1% |
| FrontierMath Tier 4 | — | 61% |
| ProofBench | — | 60% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |

## Knowledge

- DeepSeek-V3: 37.5 (#147)
- GPT-5.6 Luna: 58.6 (#31)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 67.6% | 91.6% |
| LMArena Expert | 1352 | 1479 |
| SimpleQA Verified | — | 41% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |

## Multimodal

- DeepSeek-V3: —
- GPT-5.6 Luna: 43.0 (#22)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | — | 1259 |
| Blueprint-Bench 2 | — | 22.6% |
| Furniture Assembly | — | 42.5% |
| LMArena Document | — | 1457 |

## Multilingual

- DeepSeek-V3: 48.8 (#138)
- GPT-5.6 Luna: 53.0 (#73)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | 1358 | 1418 |
| LMArena Chinese | 1390 | 1469 |
| LMArena French | 1386 | 1457 |
| LMArena German | 1374 | 1459 |
| LMArena Japanese | 1333 | 1416 |
| LMArena Korean | 1321 | 1414 |
| LMArena Russian | 1372 | 1428 |
| LMArena Spanish | 1358 | 1445 |

## Instruction Following

- DeepSeek-V3: 73.0 (#125)
- GPT-5.6 Luna: 76.0 (#48)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | 1345 | 1440 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |

## Long Context

- DeepSeek-V3: 34.0 (#248)
- GPT-5.6 Luna: 43.9 (#78)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | 1353 | 1437 |
| Fiction.LiveBench | 50% | — |

## Writing & Preference

- DeepSeek-V3: 57.6 (#124)
- GPT-5.6 Luna: 68.3 (#27)

| Benchmark | DeepSeek-V3 | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | 1375 | 1432 |
| LMArena Creative Writing | 1364 | 1397 |
| EQ-Bench Creative Writing | 1472 | 1829 |
| LMArena Multi-Turn | 1389 | 1433 |
| Short-Story Creative Writing | 77% | — |
| WildBench | 83% | — |
| EQ-Bench 4 | — | 1156 |
| LiveBench Language | 49.1% | — |

## FAQ

### Is DeepSeek-V3 better than GPT-5.6 Luna?

GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 39.2 on the Noometry Index.

### Which is cheaper, DeepSeek-V3 or GPT-5.6 Luna?

DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; GPT-5.6 Luna lists at $0.20 and $1.20.

### Is DeepSeek-V3 or GPT-5.6 Luna better for coding?

GPT-5.6 Luna scores higher on coding benchmarks: 53.9 versus 41.9 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.6 Luna does, with 1.05M tokens against 164K.

### How many benchmarks do DeepSeek-V3 and GPT-5.6 Luna share?

28 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and GPT-5.6 Luna has 51.