# DeepSeek-V3.1 vs Step 3

> DeepSeek-V3.1 is the stronger model overall, scoring 42.2 to 39.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v3-1-vs-step-3
- Last updated: 2026-10-10
- Shared benchmarks: 16

## Summary

- They share 16 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 6 categories and Step 3 in 2 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3.1 leads 43.0 to 36.2.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 53.2% for DeepSeek-V3.1 and 62.3% for Step 3.

## Snapshot

| | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| Provider | DeepSeek | StepFun |
| Noometry Index | 42.2 | 39.9 |
| Rank | 106 | 145 |
| Context | 164K | — |
| Input $/M | $0.25 | — |
| Output $/M | $0.95 | — |
| Weights | Open | Open |

## Coding

- DeepSeek-V3.1: 39.8 (#140)
- Step 3: 39.5 (#144)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Coding | 1417 | 1367 |
| WeirdML | 38.4% | — |

## Reasoning

- DeepSeek-V3.1: 27.3 (#109)
- Step 3: 27.6 (#102)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| Kagi LLM Benchmark | 53.2% | 62.3% |
| LMArena Hard Prompts | 1417 | 1355 |
| SimpleBench | 40% | — |
| DTBench | 82.7% | — |
| LMCA | 24.3% | — |
| Epoch Capabilities Index | 139.92 | — |
| ForecastBench | 58 | — |

## Math

- DeepSeek-V3.1: 36.1 (#125)
- Step 3: 34.9 (#148)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Math | 1420 | 1366 |

## Knowledge

- DeepSeek-V3.1: 43.0 (#86)
- Step 3: 36.2 (#161)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Expert | 1403 | 1333 |
| Vectara Hallucination Rate | 5.5% | — |

## Multimodal

- DeepSeek-V3.1: —
- Step 3: 34.9 (#81)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Vision | — | 1176 |

## Multilingual

- DeepSeek-V3.1: 51.8 (#100)
- Step 3: 46.5 (#154)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Non-English | 1400 | 1327 |
| LMArena Chinese | 1470 | 1396 |
| LMArena German | 1412 | 1371 |
| LMArena Korean | 1337 | 1271 |
| LMArena Russian | 1405 | 1331 |
| LMArena Spanish | 1429 | 1372 |
| LMArena French | 1447 | — |
| LMArena Japanese | 1374 | — |

## Instruction Following

- DeepSeek-V3.1: 74.1 (#105)
- Step 3: 70.6 (#159)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1332 |

## Long Context

- DeepSeek-V3.1: 36.3 (#227)
- Step 3: 40.3 (#152)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Longer Query | 1422 | 1326 |
| Fiction.LiveBench | 52.8% | — |

## Writing & Preference

- DeepSeek-V3.1: 60.5 (#93)
- Step 3: 54.5 (#146)

| Benchmark | DeepSeek-V3.1 | Step 3 |
|---|---|---|
| LMArena Text | 1420 | 1351 |
| LMArena Creative Writing | 1401 | 1322 |
| LMArena Multi-Turn | 1408 | 1340 |
| EQ-Bench Creative Writing | 1436 | — |

## FAQ

### Is DeepSeek-V3.1 better than Step 3?

DeepSeek-V3.1 is the stronger model overall, scoring 42.2 to 39.9 on the Noometry Index.

### Is DeepSeek-V3.1 or Step 3 better for coding?

They score almost the same on coding (39.8 vs 39.5); test both on your own repository before choosing.

### How many benchmarks do DeepSeek-V3.1 and Step 3 share?

16 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and Step 3 has 17.