# GPT-5.6 Luna vs Phi 3 Mini 4k Instruct

> GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 27.4 on the Noometry Index.

- Canonical page: https://noometry.com/compare/gpt-5-6-luna-vs-phi-3-mini-4k-instruct
- Last updated: 2026-10-10
- Shared benchmarks: 18

## Summary

- They share 18 benchmarks with published results for both. GPT-5.6 Luna scores higher in 8 categories and Phi 3 Mini 4k Instruct in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.5 to 24.5.
- The biggest single-benchmark swing is Chess Puzzles: 40% for GPT-5.6 Luna and 0% for Phi 3 Mini 4k Instruct.
- Phi 3 Mini 4k Instruct has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| Provider | OpenAI | Microsoft |
| Noometry Index | 54.7 | 27.4 |
| Rank | 24 | 324 |
| Context | 1.05M | — |
| Input $/M | $0.20 | — |
| Output $/M | $1.20 | — |
| Weights | Proprietary | Open |

## Coding

- GPT-5.6 Luna: 53.9 (#25)
- Phi 3 Mini 4k Instruct: 26.1 (#319)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Coding | 1466 | 1093 |
| DeepSWE | 67.2% | — |
| FrontierCode | 39.8% | — |
| CursorBench | 35.9% | — |
| LMArena WebDev | 1517 | — |
| SciCode | 53.6% | — |
| WeirdML | 60.9% | — |
| LiveBench Coding | — | 15.5% |
| ALE-Bench | 1,667 | — |
| HumanEval+ | — | 59.1% |
| MBPP+ | — | 54.2% |

## Agentic & Tool Use

- GPT-5.6 Luna: 35.3 (#37)
- Phi 3 Mini 4k Instruct: —

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| BALROG | 45.6% | — |
| GDP.pdf | 22.7% | — |
| Vending-Bench 2 | 4,095 | — |

## Reasoning

- GPT-5.6 Luna: 47.5 (#38)
- Phi 3 Mini 4k Instruct: 13.6 (#326)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| Chess Puzzles | 40% | 0% |
| LMArena Hard Prompts | 1452 | 1072 |
| ARC-AGI-2 | 59.5% | — |
| SimpleBench | 46.8% | — |
| Kagi LLM Benchmark | 49.1% | — |
| NYT Connections (extended) | 69.4% | — |
| ARC-AGI-1 | 88% | — |
| CritPt | 20.6% | — |
| LiveBench Reasoning | — | 26.8% |
| Mystery Game Puzzles | 21% | — |
| DTBench | 89.1% | — |
| LiveBench Data Analysis | — | 34.7% |
| LMCA | 48.5% | — |
| Surface Evolver Bench | 61.9% | — |
| Adversarial NLI | — | 52.8% |
| BIG-Bench Hard | — | 71.7% |
| Epoch Capabilities Index | 156.32 | — |
| HellaSwag | — | 76.7% |
| LiveBench | — | 22.4% |
| WinoGrande | — | 70.8% |

## Math

- GPT-5.6 Luna: 77.5 (#12)
- Phi 3 Mini 4k Instruct: 24.5 (#258)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Math | 1455 | 1111 |
| FrontierMath (Tiers 1-3) | 82.1% | — |
| FrontierMath Tier 4 | 61% | — |
| OTIS Mock AIME 2024-2025 | 98.3% | — |
| ProofBench | 60% | — |
| LiveBench Math | — | 15.7% |

## Knowledge

- GPT-5.6 Luna: 58.6 (#31)
- Phi 3 Mini 4k Instruct: 27.9 (#241)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Expert | 1479 | 1045 |
| GPQA Diamond | 91.6% | — |
| SimpleQA Verified | 41% | — |
| ARC (AI2) Challenge | — | 84.9% |
| MMLU | — | 68.8% |
| OpenBookQA | — | 88% |
| TriviaQA | — | 64% |

## Multimodal

- GPT-5.6 Luna: 43.0 (#22)
- Phi 3 Mini 4k Instruct: —

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Vision | 1259 | — |
| Blueprint-Bench 2 | 22.6% | — |
| Furniture Assembly | 42.5% | — |
| LMArena Document | 1457 | — |

## Multilingual

- GPT-5.6 Luna: 53.0 (#73)
- Phi 3 Mini 4k Instruct: 26.5 (#275)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Non-English | 1418 | 1021 |
| LMArena Chinese | 1469 | 1021 |
| LMArena French | 1457 | 1075 |
| LMArena German | 1459 | 1044 |
| LMArena Japanese | 1416 | 935 |
| LMArena Korean | 1414 | 905 |
| LMArena Russian | 1428 | 1022 |
| LMArena Spanish | 1445 | 1084 |

## Instruction Following

- GPT-5.6 Luna: 76.0 (#48)
- Phi 3 Mini 4k Instruct: 48.0 (#298)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Instruction Following | 1440 | 1053 |
| LiveBench Instruction Following | — | 39.1% |

## Long Context

- GPT-5.6 Luna: 43.9 (#78)
- Phi 3 Mini 4k Instruct: 31.7 (#271)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Longer Query | 1437 | 1044 |

## Writing & Preference

- GPT-5.6 Luna: 68.3 (#27)
- Phi 3 Mini 4k Instruct: 27.7 (#295)

| Benchmark | GPT-5.6 Luna | Phi 3 Mini 4k Instruct |
|---|---|---|
| LMArena Text | 1432 | 1073 |
| LMArena Creative Writing | 1397 | 1037 |
| LMArena Multi-Turn | 1433 | 1018 |
| EQ-Bench Creative Writing | 1829 | — |
| EQ-Bench 4 | 1156 | — |
| LiveBench Language | — | 9.2% |

## FAQ

### Is GPT-5.6 Luna better than Phi 3 Mini 4k Instruct?

GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 27.4 on the Noometry Index.

### Is GPT-5.6 Luna or Phi 3 Mini 4k Instruct better for coding?

GPT-5.6 Luna scores higher on coding benchmarks: 53.9 versus 26.1 in the Noometry coding category.

### How many benchmarks do GPT-5.6 Luna and Phi 3 Mini 4k Instruct share?

18 benchmarks have published results for both models. GPT-5.6 Luna has 51 scored results on Noometry and Phi 3 Mini 4k Instruct has 35.