# GPT-6 Luna vs Step 3.5 Flash

> GPT-6 Luna is the stronger model overall, scoring 53.3 to 42.3 on the Noometry Index.

- Canonical page: https://noometry.com/compare/gpt-6-luna-vs-step-3-5-flash
- Last updated: 2026-10-10
- Shared benchmarks: 18

## Summary

- They share 18 benchmarks with published results for both. GPT-6 Luna scores higher in 7 categories and Step 3.5 Flash in 1 category; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-6 Luna leads 76.1 to 42.6.
- The biggest single-benchmark swing is NYT Connections (extended): 68.7% for GPT-6 Luna and 28.4% for Step 3.5 Flash.
- Step 3.5 Flash is cheaper at $0.10 / $0.30 per million input/output tokens, against $0.10 / $0.50 for GPT-6 Luna.
- GPT-6 Luna accepts more context: 1.05M tokens versus 256K.
- Step 3.5 Flash has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| Provider | OpenAI | StepFun |
| Noometry Index | 53.3 | 42.3 |
| Rank | 36 | 116 |
| Context | 1.05M | 256K |
| Input $/M | $0.10 | $0.10 |
| Output $/M | $0.50 | $0.30 |
| Weights | Proprietary | Open |

## Coding

- GPT-6 Luna: 55.5 (#25)
- Step 3.5 Flash: 42.4 (#105)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Coding | 1439 | 1436 |
| DeepSWE | 66.6% | — |
| FrontierCode | 42.4% | — |
| LMArena WebDev | 1581 | — |
| SciCode | 54.6% | — |
| ALE-Bench | 1,577 | — |

## Agentic & Tool Use

- GPT-6 Luna: 33.3 (#54)
- Step 3.5 Flash: —

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| APEX-Agents | 44.3% | — |
| GDP.pdf | 23% | — |

## Reasoning

- GPT-6 Luna: 48.2 (#41)
- Step 3.5 Flash: 22.2 (#202)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| NYT Connections (extended) | 68.7% | 28.4% |
| LMArena Hard Prompts | 1411 | 1411 |
| ARC-AGI-2 | 59.3% | — |
| ARC-AGI-1 | 86.7% | — |
| CritPt | 19.4% | — |
| Chess Puzzles | 31% | — |
| Mystery Game Puzzles | 7% | — |
| DTBench | 90.1% | — |
| LMCA | 44.5% | — |
| Epoch Capabilities Index | 156.28 | — |

## Math

- GPT-6 Luna: 76.1 (#15)
- Step 3.5 Flash: 42.6 (#84)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Math | 1416 | 1408 |
| FrontierMath (Tiers 1-3) | 78.9% | — |
| FrontierMath Tier 4 | 56.1% | — |
| MathArena Final-Answer Competitions | — | 66.8% |
| OTIS Mock AIME 2024-2025 | 98.9% | — |
| ProofBench | 64% | — |

## Knowledge

- GPT-6 Luna: 57.0 (#41)
- Step 3.5 Flash: 39.6 (#132)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Expert | 1444 | 1421 |
| GPQA Diamond | 90.5% | — |
| SimpleQA Verified | 41.4% | — |

## Multimodal

- GPT-6 Luna: 42.4 (#30)
- Step 3.5 Flash: —

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Vision | 1217 | — |
| Blueprint-Bench 2 | 31.2% | — |
| Furniture Assembly | 44.2% | — |

## Multilingual

- GPT-6 Luna: 50.5 (#117)
- Step 3.5 Flash: 50.5 (#119)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Non-English | 1386 | 1385 |
| LMArena Chinese | 1433 | 1447 |
| LMArena French | 1420 | 1421 |
| LMArena German | 1369 | 1405 |
| LMArena Japanese | 1369 | 1354 |
| LMArena Korean | 1360 | 1352 |
| LMArena Russian | 1394 | 1385 |
| LMArena Spanish | 1393 | 1419 |

## Instruction Following

- GPT-6 Luna: 74.3 (#99)
- Step 3.5 Flash: 73.1 (#124)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Instruction Following | 1409 | 1385 |

## Long Context

- GPT-6 Luna: 43.0 (#111)
- Step 3.5 Flash: 42.8 (#117)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Longer Query | 1409 | 1402 |

## Writing & Preference

- GPT-6 Luna: 58.3 (#119)
- Step 3.5 Flash: 58.8 (#113)

| Benchmark | GPT-6 Luna | Step 3.5 Flash |
|---|---|---|
| LMArena Text | 1391 | 1403 |
| LMArena Creative Writing | 1363 | 1357 |
| LMArena Multi-Turn | 1396 | 1405 |

## FAQ

### Is GPT-6 Luna better than Step 3.5 Flash?

GPT-6 Luna is the stronger model overall, scoring 53.3 to 42.3 on the Noometry Index.

### Which is cheaper, GPT-6 Luna or Step 3.5 Flash?

Step 3.5 Flash is cheaper. It lists at $0.10 per million input tokens and $0.30 per million output tokens; GPT-6 Luna lists at $0.10 and $0.50.

### Is GPT-6 Luna or Step 3.5 Flash better for coding?

GPT-6 Luna scores higher on coding benchmarks: 55.5 versus 42.4 in the Noometry coding category.

### Which has the bigger context window?

GPT-6 Luna does, with 1.05M tokens against 256K.

### How many benchmarks do GPT-6 Luna and Step 3.5 Flash share?

18 benchmarks have published results for both models. GPT-6 Luna has 42 scored results on Noometry and Step 3.5 Flash has 19.