# Hy3 vs Muse Spark

> Muse Spark is the stronger model overall, scoring 50.6 to 44.2 on the Noometry Index.

- Canonical page: https://noometry.com/compare/hy3-vs-muse-spark
- Last updated: 2026-10-10
- Shared benchmarks: 16

## Summary

- They share 16 benchmarks with published results for both. Hy3 scores higher in 1 category and Muse Spark in 7 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Muse Spark leads 65.7 to 40.8.
- Hy3 has downloadable open weights; the other is API-only.

## Snapshot

| | Hy3 | Muse Spark |
|---|---|---|
| Provider | Tencent | Meta |
| Noometry Index | 44.2 | 50.6 |
| Rank | 79 | 46 |
| Context | 262K | — |
| Input $/M | $0.0825 | — |
| Output $/M | $0.33 | — |
| Weights | Open | Proprietary |

## Coding

- Hy3: 46.8 (#63)
- Muse Spark: 46.2 (#69)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Coding | 1464 | 1481 |
| LMArena WebDev | 1508 | — |
| SciCode | — | 51.5% |

## Reasoning

- Hy3: 26.1 (#136)
- Muse Spark: 35.9 (#67)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Hard Prompts | 1447 | 1474 |
| NYT Connections (extended) | 41.2% | — |
| CritPt | — | 11.3% |
| Epoch Capabilities Index | — | 152.04 |

## Math

- Hy3: 40.1 (#93)
- Muse Spark: 47.8 (#66)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Math | 1475 | 1455 |
| OTIS Mock AIME 2024-2025 | — | 88.9% |
| ProofBench | — | 17% |
| FrontierMath (Feb 2025 set) | — | 39% |
| FrontierMath Tier 4 (v1) | — | 14.6% |

## Knowledge

- Hy3: 40.8 (#114)
- Muse Spark: 65.7 (#13)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Expert | 1460 | 1457 |
| GPQA Diamond | — | 89.8% |
| Humanity's Last Exam | — | 40.6% |

## Multimodal

- Hy3: —
- Muse Spark: 43.4 (#24)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Vision | — | 1306 |
| LMArena Document | — | 1444 |

## Multilingual

- Hy3: 53.5 (#65)
- Muse Spark: 56.1 (#24)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Non-English | 1426 | 1464 |
| LMArena Chinese | 1493 | 1509 |
| LMArena French | 1461 | 1497 |
| LMArena German | 1439 | 1497 |
| LMArena Korean | 1395 | 1459 |
| LMArena Russian | 1432 | 1466 |
| LMArena Spanish | 1456 | 1472 |
| LMArena Japanese | 1392 | — |

## Instruction Following

- Hy3: 75.1 (#70)
- Muse Spark: 75.9 (#51)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Instruction Following | 1426 | 1442 |

## Long Context

- Hy3: 44.1 (#75)
- Muse Spark: 44.4 (#69)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Longer Query | 1442 | 1451 |

## Writing & Preference

- Hy3: 62.2 (#81)
- Muse Spark: 66.0 (#39)

| Benchmark | Hy3 | Muse Spark |
|---|---|---|
| LMArena Text | 1439 | 1474 |
| LMArena Creative Writing | 1402 | 1459 |
| LMArena Multi-Turn | 1436 | 1477 |

## FAQ

### Is Hy3 better than Muse Spark?

Muse Spark is the stronger model overall, scoring 50.6 to 44.2 on the Noometry Index.

### Is Hy3 or Muse Spark better for coding?

They score almost the same on coding (46.8 vs 46.2); test both on your own repository before choosing.

### How many benchmarks do Hy3 and Muse Spark share?

16 benchmarks have published results for both models. Hy3 has 19 scored results on Noometry and Muse Spark has 27.