# GPT-5.6 Luna vs Hy3

> GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 43.6 on the Noometry Index. Hy3 costs 2.0× less per token, which makes it the better buy when GPT-5.6 Luna's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-5-6-luna-vs-hy3
- Last updated: 2026-10-10
- Shared benchmarks: 17

## Summary

- They share 17 benchmarks with published results for both. GPT-5.6 Luna scores higher in 6 categories and Hy3 in 2 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.5 to 37.4.
- The biggest single-benchmark swing is NYT Connections (extended): 69.4% for GPT-5.6 Luna and 41.2% for Hy3.
- Hy3 is cheaper at $0.13 / $0.53 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 262K.
- Hy3 has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-5.6 Luna | Hy3 |
|---|---|---|
| Provider | OpenAI | Tencent |
| Noometry Index | 54.7 | 43.6 |
| Rank | 24 | 79 |
| Context | 1.05M | 262K |
| Input $/M | $0.20 | $0.13 |
| Output $/M | $1.20 | $0.53 |
| Weights | Proprietary | Open |

## Coding

- GPT-5.6 Luna: 53.9 (#25)
- Hy3: 46.6 (#60)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena WebDev | 1517 | 1509 |
| LMArena Coding | 1466 | 1465 |
| DeepSWE | 67.2% | — |
| FrontierCode | 39.8% | — |
| CursorBench | 35.9% | — |
| SciCode | 53.6% | — |
| WeirdML | 60.9% | — |
| ALE-Bench | 1,667 | — |

## Agentic & Tool Use

- GPT-5.6 Luna: 35.3 (#37)
- Hy3: —

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| BALROG | 45.6% | — |
| GDP.pdf | 22.7% | — |
| Vending-Bench 2 | 4,095 | — |

## Reasoning

- GPT-5.6 Luna: 47.5 (#38)
- Hy3: 25.0 (#134)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| NYT Connections (extended) | 69.4% | 41.2% |
| LMArena Hard Prompts | 1452 | 1449 |
| ARC-AGI-2 | 59.5% | — |
| SimpleBench | 46.8% | — |
| Kagi LLM Benchmark | 49.1% | — |
| ARC-AGI-1 | 88% | — |
| CritPt | 20.6% | — |
| Chess Puzzles | 40% | — |
| Mystery Game Puzzles | 21% | — |
| DTBench | 89.1% | — |
| LMCA | 48.5% | — |
| Surface Evolver Bench | 61.9% | — |
| Epoch Capabilities Index | 156.32 | — |

## Math

- GPT-5.6 Luna: 77.5 (#12)
- Hy3: 37.4 (#98)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena Math | 1455 | 1475 |
| FrontierMath (Tiers 1-3) | 82.1% | — |
| FrontierMath Tier 4 | 61% | — |
| OTIS Mock AIME 2024-2025 | 98.3% | — |
| ProofBench | 60% | — |

## Knowledge

- GPT-5.6 Luna: 58.6 (#31)
- Hy3: 40.2 (#108)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena Expert | 1479 | 1461 |
| GPQA Diamond | 91.6% | — |
| SimpleQA Verified | 41% | — |

## Multimodal

- GPT-5.6 Luna: 43.0 (#22)
- Hy3: —

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena Vision | 1259 | — |
| Blueprint-Bench 2 | 22.6% | — |
| Furniture Assembly | 42.5% | — |
| LMArena Document | 1457 | — |

## Multilingual

- GPT-5.6 Luna: 53.0 (#73)
- Hy3: 53.7 (#60)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena Non-English | 1418 | 1427 |
| LMArena Chinese | 1469 | 1496 |
| LMArena French | 1457 | 1461 |
| LMArena Korean | 1414 | 1390 |
| LMArena Russian | 1428 | 1435 |
| LMArena Spanish | 1445 | 1448 |
| LMArena German | 1459 | — |
| LMArena Japanese | 1416 | — |

## Instruction Following

- GPT-5.6 Luna: 76.0 (#48)
- Hy3: 75.3 (#68)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena Instruction Following | 1440 | 1425 |

## Long Context

- GPT-5.6 Luna: 43.9 (#78)
- Hy3: 44.1 (#70)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena Longer Query | 1437 | 1443 |

## Writing & Preference

- GPT-5.6 Luna: 68.3 (#27)
- Hy3: 62.5 (#75)

| Benchmark | GPT-5.6 Luna | Hy3 |
|---|---|---|
| LMArena Text | 1432 | 1442 |
| LMArena Creative Writing | 1397 | 1399 |
| LMArena Multi-Turn | 1433 | 1436 |
| EQ-Bench Creative Writing | 1829 | — |
| EQ-Bench 4 | 1156 | — |

## FAQ

### Is GPT-5.6 Luna better than Hy3?

GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 43.6 on the Noometry Index. Hy3 costs 2.0× less per token, which makes it the better buy when GPT-5.6 Luna's lead doesn't matter for your workload.

### Which is cheaper, GPT-5.6 Luna or Hy3?

Hy3 is cheaper. It lists at $0.13 per million input tokens and $0.53 per million output tokens; GPT-5.6 Luna lists at $0.20 and $1.20.

### Is GPT-5.6 Luna or Hy3 better for coding?

GPT-5.6 Luna scores higher on coding benchmarks: 53.9 versus 46.6 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.6 Luna does, with 1.05M tokens against 262K.

### How many benchmarks do GPT-5.6 Luna and Hy3 share?

17 benchmarks have published results for both models. GPT-5.6 Luna has 51 scored results on Noometry and Hy3 has 17.