# GPT-5.4 Pro vs Step 3

> GPT-5.4 Pro is the stronger model overall, scoring 58.9 to 40.5 on the Noometry Index.

- Canonical page: https://noometry.com/compare/gpt-5-4-pro-vs-step-3
- Last updated: 2026-10-11
- Shared benchmarks: 0

## Summary

- The widest gap is in reasoning, where GPT-5.4 Pro leads 70.7 to 28.4.
- Step 3 has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-5.4 Pro | Step 3 |
|---|---|---|
| Provider | OpenAI | StepFun |
| Noometry Index | 58.9 | 40.5 |
| Rank | 18 | 149 |
| Context | 1.05M | — |
| Input $/M | $30 | — |
| Output $/M | $180 | — |
| Weights | Proprietary | Open |

## Coding

- GPT-5.4 Pro: 43.2 (#90)
- Step 3: 40.1 (#147)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| WeirdML | 57.4% | — |
| LMArena Coding | — | 1367 |

## Reasoning

- GPT-5.4 Pro: 70.7 (#13)
- Step 3: 28.4 (#105)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| ARC-AGI-2 | 83.3% | — |
| SimpleBench | 74.1% | — |
| Kagi LLM Benchmark | — | 62.3% |
| ARC-AGI-1 | 94.5% | — |
| CritPt | 30% | — |
| Chess Puzzles | 58.6% | — |
| EnigmaEval | 23.8% | — |
| LMArena Hard Prompts | — | 1355 |
| Epoch Capabilities Index | 158.93 | — |

## Math

- GPT-5.4 Pro: 72.4 (#22)
- Step 3: 37.6 (#148)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 82.5% | — |
| FrontierMath Tier 4 | 58.5% | — |
| LMArena Math | — | 1366 |
| FrontierMath (Feb 2025 set) | 50% | — |
| FrontierMath Tier 4 (v1) | 37.5% | — |

## Knowledge

- GPT-5.4 Pro: 68.3 (#7)
- Step 3: 36.8 (#164)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| GPQA Diamond | 94.6% | — |
| Humanity's Last Exam | 44.3% | — |
| SimpleQA Verified | 46.3% | — |
| Vectara Hallucination Rate | 8.3% | — |
| LMArena Expert | — | 1333 |

## Multimodal

- GPT-5.4 Pro: —
- Step 3: 35.5 (#86)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| LMArena Vision | — | 1177 |

## Multilingual

- GPT-5.4 Pro: —
- Step 3: 46.3 (#159)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| LMArena Non-English | — | 1327 |
| LMArena Chinese | — | 1397 |
| LMArena German | — | 1371 |
| LMArena Korean | — | 1269 |
| LMArena Russian | — | 1331 |
| LMArena Spanish | — | 1371 |

## Instruction Following

- GPT-5.4 Pro: —
- Step 3: 70.4 (#164)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| LMArena Instruction Following | — | 1332 |

## Long Context

- GPT-5.4 Pro: —
- Step 3: 40.3 (#157)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| LMArena Longer Query | — | 1326 |

## Writing & Preference

- GPT-5.4 Pro: —
- Step 3: 54.3 (#151)

| Benchmark | GPT-5.4 Pro | Step 3 |
|---|---|---|
| LMArena Text | — | 1350 |
| LMArena Creative Writing | — | 1321 |
| LMArena Multi-Turn | — | 1341 |

## FAQ

### Is GPT-5.4 Pro better than Step 3?

GPT-5.4 Pro is the stronger model overall, scoring 58.9 to 40.5 on the Noometry Index.

### Is GPT-5.4 Pro or Step 3 better for coding?

GPT-5.4 Pro scores higher on coding benchmarks: 43.2 versus 40.1 in the Noometry coding category.

### How many benchmarks do GPT-5.4 Pro and Step 3 share?

0 benchmarks have published results for both models. GPT-5.4 Pro has 16 scored results on Noometry and Step 3 has 17.