# GPT-5.6 Terra vs Hy4 preview

> GPT-5.6 Terra is the stronger model overall, scoring 59.2 to 45.3 on the Noometry Index. Hy4 preview costs 3.6× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-5-6-terra-vs-hy4-preview
- Last updated: 2026-10-12
- Shared benchmarks: 3

## Summary

- They share 3 benchmarks with published results for both. GPT-5.6 Terra scores higher in 3 categories and Hy4 preview in 0 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.6 Terra leads 60.7 to 31.9.
- The biggest single-benchmark swing is NYT Connections (extended): 78.4% for GPT-5.6 Terra and 68.2% for Hy4 preview.
- Hy4 preview is cheaper at $0.83 / $2.50 per million input/output tokens, against $2 / $12 for GPT-5.6 Terra.
- GPT-5.6 Terra accepts more context: 1.05M tokens versus 1.05M.
- Hy4 preview has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| Provider | OpenAI | Tencent |
| Noometry Index | 59.2 | 45.3 |
| Rank | 17 | 73 |
| Context | 1.05M | 1.05M |
| Input $/M | $2 | $0.83 |
| Output $/M | $12 | $2.50 |
| Weights | Proprietary | Open |

## Coding

- GPT-5.6 Terra: 57.7 (#19)
- Hy4 preview: 51.6 (#38)

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| LMArena WebDev | 1522 | 1632 |
| DeepSWE | 69.6% | — |
| FrontierCode | 41.3% | — |
| CursorBench | 41.3% | — |
| SciCode | 55% | — |
| WeirdML | 78.3% | — |
| LMArena Coding | 1484 | — |
| ALE-Bench | 1,951 | — |

## Agentic & Tool Use

- GPT-5.6 Terra: 40.1 (#25)
- Hy4 preview: —

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| APEX-Agents | 58.2% | — |
| BALROG | 53.2% | — |
| GDP.pdf | 24.7% | — |
| Vending-Bench 2 | 7,343 | — |

## Reasoning

- GPT-5.6 Terra: 60.7 (#21)
- Hy4 preview: 31.9 (#79)

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| NYT Connections (extended) | 78.4% | 68.2% |
| ARC-AGI-2 | 83.9% | — |
| SimpleBench | 48.9% | — |
| Kagi LLM Benchmark | 51.3% | — |
| ARC-AGI-1 | 96.5% | — |
| CritPt | 30% | — |
| Chess Puzzles | 54% | — |
| LMArena Hard Prompts | 1468 | — |
| Mystery Game Puzzles | 35% | — |
| DTBench | 93.3% | — |
| LMCA | 55% | — |
| Surface Evolver Bench | 83.8% | — |
| Epoch Capabilities Index | 159.62 | — |

## Math

- GPT-5.6 Terra: 81.6 (#12)
- Hy4 preview: 55.7 (#42)

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| ProofBench | 74% | 75% |
| FrontierMath (Tiers 1-3) | 86% | — |
| FrontierMath Tier 4 | 70.7% | — |
| OTIS Mock AIME 2024-2025 | 99.7% | — |
| LMArena Math | 1466 | — |

## Knowledge

- GPT-5.6 Terra: 61.2 (#30)
- Hy4 preview: —

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| GPQA Diamond | 93.3% | — |
| SimpleQA Verified | 43.2% | — |
| LMArena Expert | 1492 | — |

## Multimodal

- GPT-5.6 Terra: 47.3 (#11)
- Hy4 preview: —

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| LMArena Vision | 1271 | — |
| Blueprint-Bench 2 | 30.8% | — |
| Furniture Assembly | 54.2% | — |
| LMArena Document | 1472 | — |

## Multilingual

- GPT-5.6 Terra: 54.4 (#44)
- Hy4 preview: —

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| LMArena Non-English | 1439 | — |
| LMArena Chinese | 1513 | — |
| LMArena French | 1471 | — |
| LMArena German | 1460 | — |
| LMArena Japanese | 1457 | — |
| LMArena Korean | 1425 | — |
| LMArena Russian | 1450 | — |
| LMArena Spanish | 1448 | — |

## Instruction Following

- GPT-5.6 Terra: 76.4 (#40)
- Hy4 preview: —

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| LMArena Instruction Following | 1454 | — |

## Long Context

- GPT-5.6 Terra: 44.4 (#68)
- Hy4 preview: —

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| LMArena Longer Query | 1451 | — |

## Writing & Preference

- GPT-5.6 Terra: 70.2 (#23)
- Hy4 preview: —

| Benchmark | GPT-5.6 Terra | Hy4 preview |
|---|---|---|
| LMArena Text | 1447 | — |
| LMArena Creative Writing | 1410 | — |
| EQ-Bench Creative Writing | 1855 | — |
| EQ-Bench 4 | 1234 | — |
| LMArena Multi-Turn | 1449 | — |

## FAQ

### Is GPT-5.6 Terra better than Hy4 preview?

GPT-5.6 Terra is the stronger model overall, scoring 59.2 to 45.3 on the Noometry Index. Hy4 preview costs 3.6× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.

### Which is cheaper, GPT-5.6 Terra or Hy4 preview?

Hy4 preview is cheaper. It lists at $0.83 per million input tokens and $2.50 per million output tokens; GPT-5.6 Terra lists at $2 and $12.

### Is GPT-5.6 Terra or Hy4 preview better for coding?

GPT-5.6 Terra scores higher on coding benchmarks: 57.7 versus 51.6 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.6 Terra does, with 1.05M tokens against 1.05M.

### How many benchmarks do GPT-5.6 Terra and Hy4 preview share?

3 benchmarks have published results for both models. GPT-5.6 Terra has 52 scored results on Noometry and Hy4 preview has 3.