# GPT-6 Sol vs Step 3.7 Flash

> GPT-6 Sol is the stronger model overall, scoring 61.8 to 37.3 on the Noometry Index. Step 3.7 Flash costs 9.6× less per token, which makes it the better buy when GPT-6 Sol's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-6-sol-vs-step-3-7-flash
- Last updated: 2026-10-11
- Shared benchmarks: 4

## Summary

- They share 4 benchmarks with published results for both. GPT-6 Sol 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-6 Sol leads 74.0 to 21.6.
- The biggest single-benchmark swing is NYT Connections (extended): 90.1% for GPT-6 Sol 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 $2 / $10 for GPT-6 Sol.
- GPT-6 Sol accepts more context: 1.05M tokens versus 256K.
- Step 3.7 Flash has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| Provider | OpenAI | StepFun |
| Noometry Index | 61.8 | 37.3 |
| Rank | 12 | 207 |
| Context | 1.05M | 256K |
| Input $/M | $2 | $0.18 |
| Output $/M | $10 | $1.11 |
| Weights | Proprietary | Open |

## Coding

- GPT-6 Sol: 60.1 (#11)
- Step 3.7 Flash: 40.0 (#150)

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| SciCode | 57.6% | 40% |
| ALE-Bench | 2,462 | 694.12 |
| DeepSWE | 68.8% | — |
| FrontierCode | 49.3% | — |
| LMArena WebDev | 1688 | — |
| LMArena Coding | 1447 | — |

## Agentic & Tool Use

- GPT-6 Sol: 37.2 (#36)
- Step 3.7 Flash: —

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| APEX-Agents | 54.3% | — |
| GDP.pdf | 26.4% | — |
| Vending-Bench 2 | 14,428 | — |

## Reasoning

- GPT-6 Sol: 74.0 (#9)
- Step 3.7 Flash: 21.6 (#219)

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| NYT Connections (extended) | 90.1% | 39.7% |
| CritPt | 30.9% | 2.3% |
| ARC-AGI-2 | 89.6% | — |
| ARC-AGI-1 | 95.5% | — |
| EBR-Bench | 53.3% | — |
| LMArena Hard Prompts | 1418 | — |
| Mystery Game Puzzles | 56% | — |
| DTBench | 97.3% | — |
| LMCA | 59.1% | — |
| Epoch Capabilities Index | 162.72 | — |

## Math

- GPT-6 Sol: 87.2 (#7)
- Step 3.7 Flash: 42.9 (#82)

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| FrontierMath (Tiers 1-3) | 89.8% | — |
| FrontierMath Tier 4 | 90% | — |
| MathArena Final-Answer Competitions | — | 68.5% |
| OTIS Mock AIME 2024-2025 | 100% | — |
| ProofBench | 83% | — |
| LMArena Math | 1402 | — |

## Knowledge

- GPT-6 Sol: 64.8 (#15)
- Step 3.7 Flash: —

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| GPQA Diamond | 94.3% | — |
| SimpleQA Verified | 60.7% | — |
| Vectara Hallucination Rate | 6.5% | — |
| LMArena Expert | 1439 | — |

## Multimodal

- GPT-6 Sol: 47.6 (#10)
- Step 3.7 Flash: —

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| LMArena Vision | 1245 | — |
| Blueprint-Bench 2 | 36.9% | — |
| Furniture Assembly | 58.3% | — |

## Multilingual

- GPT-6 Sol: 50.5 (#118)
- Step 3.7 Flash: —

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| LMArena Non-English | 1385 | — |
| LMArena Chinese | 1405 | — |
| LMArena French | 1410 | — |
| LMArena German | 1390 | — |
| LMArena Japanese | 1385 | — |
| LMArena Korean | 1341 | — |
| LMArena Russian | 1401 | — |
| LMArena Spanish | 1384 | — |

## Instruction Following

- GPT-6 Sol: 74.5 (#94)
- Step 3.7 Flash: —

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| LMArena Instruction Following | 1412 | — |

## Long Context

- GPT-6 Sol: 43.1 (#108)
- Step 3.7 Flash: —

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| LMArena Longer Query | 1411 | — |

## Writing & Preference

- GPT-6 Sol: 71.9 (#18)
- Step 3.7 Flash: —

| Benchmark | GPT-6 Sol | Step 3.7 Flash |
|---|---|---|
| LMArena Text | 1395 | — |
| LMArena Creative Writing | 1378 | — |
| EQ-Bench Creative Writing | 2125 | — |
| LMArena Multi-Turn | 1412 | — |

## FAQ

### Is GPT-6 Sol better than Step 3.7 Flash?

GPT-6 Sol is the stronger model overall, scoring 61.8 to 37.3 on the Noometry Index. Step 3.7 Flash costs 9.6× less per token, which makes it the better buy when GPT-6 Sol's lead doesn't matter for your workload.

### Which is cheaper, GPT-6 Sol 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-6 Sol lists at $2 and $10.

### Is GPT-6 Sol or Step 3.7 Flash better for coding?

GPT-6 Sol scores higher on coding benchmarks: 60.1 versus 40.0 in the Noometry coding category.

### Which has the bigger context window?

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

### How many benchmarks do GPT-6 Sol and Step 3.7 Flash share?

4 benchmarks have published results for both models. GPT-6 Sol has 45 scored results on Noometry and Step 3.7 Flash has 5.