# GPT-5.4 mini vs Step 3.7 Flash

> GPT-5.4 mini is the stronger model overall, scoring 45.0 to 37.3 on the Noometry Index. Step 3.7 Flash costs 4.1× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-5-4-mini-vs-step-3-7-flash
- Last updated: 2026-10-11
- Shared benchmarks: 4

## Summary

- They share 4 benchmarks with published results for both. GPT-5.4 mini scores higher in 3 categories and Step 3.7 Flash in 0 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 mini leads 30.4 to 21.6.
- The biggest single-benchmark swing is NYT Connections (extended): 61.8% for GPT-5.4 mini and 39.7% for Step 3.7 Flash.
- Step 3.7 Flash is cheaper at $0.18 / $1.11 per million input/output tokens, against $0.75 / $4.50 for GPT-5.4 mini.
- GPT-5.4 mini accepts more context: 400K tokens versus 256K.
- Step 3.7 Flash has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| Provider | OpenAI | StepFun |
| Noometry Index | 45.0 | 37.3 |
| Rank | 76 | 207 |
| Context | 400K | 256K |
| Input $/M | $0.75 | $0.18 |
| Output $/M | $4.50 | $1.11 |
| Weights | Proprietary | Open |

## Coding

- GPT-5.4 mini: 45.2 (#72)
- Step 3.7 Flash: 40.0 (#150)

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| SciCode | 49.9% | 40% |
| ALE-Bench | 1,189 | 694.12 |
| FrontierCode | 27% | — |
| LMArena WebDev | 1397 | — |
| WeirdML | 60.3% | — |
| LMArena Coding | 1438 | — |

## Agentic & Tool Use

- GPT-5.4 mini: 29.9 (#81)
- Step 3.7 Flash: —

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| DeepResearch Bench | 36.3% | — |

## Reasoning

- GPT-5.4 mini: 30.4 (#85)
- Step 3.7 Flash: 21.6 (#219)

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| NYT Connections (extended) | 61.8% | 39.7% |
| CritPt | 10% | 2.3% |
| ARC-AGI-2 | 18.9% | — |
| Kagi LLM Benchmark | 37.9% | — |
| ARC-AGI-1 | 63.7% | — |
| Chess Puzzles | 24% | — |
| Thematic Generalization | 61.7% | — |
| LMArena Hard Prompts | 1424 | — |
| Mystery Game Puzzles | 11% | — |
| DTBench | 80% | — |
| LMCA | 40.8% | — |
| Epoch Capabilities Index | 148.84 | — |
| ForecastBench | 57 | — |

## Math

- GPT-5.4 mini: 45.5 (#75)
- Step 3.7 Flash: 42.9 (#82)

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| FrontierMath (Tiers 1-3) | 51.2% | — |
| FrontierMath Tier 4 | 9.8% | — |
| MathArena Final-Answer Competitions | — | 68.5% |
| OTIS Mock AIME 2024-2025 | 88.9% | — |
| ProofBench | 21% | — |
| LMArena Math | 1419 | — |
| FrontierMath (Feb 2025 set) | 28.3% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |

## Knowledge

- GPT-5.4 mini: 51.5 (#67)
- Step 3.7 Flash: —

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| GPQA Diamond | 86.9% | — |
| SimpleQA Verified | 29.4% | — |
| Vectara Hallucination Rate | 5.5% | — |
| LMArena Expert | 1435 | — |

## Multimodal

- GPT-5.4 mini: 39.7 (#56)
- Step 3.7 Flash: —

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| LMArena Vision | 1245 | — |

## Multilingual

- GPT-5.4 mini: 51.9 (#96)
- Step 3.7 Flash: —

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| LMArena Non-English | 1405 | — |
| LMArena Chinese | 1446 | — |
| LMArena French | 1440 | — |
| LMArena German | 1409 | — |
| LMArena Japanese | 1374 | — |
| LMArena Korean | 1368 | — |
| LMArena Russian | 1417 | — |
| LMArena Spanish | 1405 | — |

## Instruction Following

- GPT-5.4 mini: 74.1 (#102)
- Step 3.7 Flash: —

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| LMArena Instruction Following | 1405 | — |

## Long Context

- GPT-5.4 mini: 43.0 (#112)
- Step 3.7 Flash: —

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| LMArena Longer Query | 1407 | — |

## Writing & Preference

- GPT-5.4 mini: 64.0 (#58)
- Step 3.7 Flash: —

| Benchmark | GPT-5.4 mini | Step 3.7 Flash |
|---|---|---|
| LMArena Text | 1412 | — |
| LMArena Creative Writing | 1370 | — |
| EQ-Bench Creative Writing | 1665 | — |
| LMArena Multi-Turn | 1429 | — |

## FAQ

### Is GPT-5.4 mini better than Step 3.7 Flash?

GPT-5.4 mini is the stronger model overall, scoring 45.0 to 37.3 on the Noometry Index. Step 3.7 Flash costs 4.1× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.

### Which is cheaper, GPT-5.4 mini or Step 3.7 Flash?

Step 3.7 Flash is cheaper. It lists at $0.18 per million input tokens and $1.11 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.

### Is GPT-5.4 mini or Step 3.7 Flash better for coding?

GPT-5.4 mini scores higher on coding benchmarks: 45.2 versus 40.0 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.4 mini does, with 400K tokens against 256K.

### How many benchmarks do GPT-5.4 mini and Step 3.7 Flash share?

4 benchmarks have published results for both models. GPT-5.4 mini has 46 scored results on Noometry and Step 3.7 Flash has 5.