# DeepSeek-V3.2-Exp vs Step 3.5 Flash

> DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 41.7 on the Noometry Index. Step 3.5 Flash costs 1.9× less per token, which makes it the better buy when DeepSeek-V3.2-Exp's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/deepseek-v3-2-exp-vs-step-3-5-flash
- Last updated: 2026-10-10
- Shared benchmarks: 19

## Summary

- They share 19 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 8 categories and Step 3.5 Flash in 0 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3.2-Exp leads 51.6 to 39.0.
- The biggest single-benchmark swing is MathArena Final-Answer Competitions: 57.7% for DeepSeek-V3.2-Exp and 66.8% for Step 3.5 Flash.
- Step 3.5 Flash is cheaper at $0.10 / $0.30 per million input/output tokens, against $0.26 / $0.38 for DeepSeek-V3.2-Exp.
- Step 3.5 Flash accepts more context: 256K tokens versus 164K.

## Snapshot

| | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| Provider | DeepSeek | StepFun |
| Noometry Index | 44.9 | 41.7 |
| Rank | 68 | 116 |
| Context | 164K | 256K |
| Input $/M | $0.26 | $0.10 |
| Output $/M | $0.38 | $0.30 |
| Weights | Open | Open |

## Coding

- DeepSeek-V3.2-Exp: 46.4 (#61)
- Step 3.5 Flash: 41.7 (#106)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| LMArena Coding | 1454 | 1436 |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| LMArena WebDev | 1362 | — |
| SWE-bench Multilingual | 59% | — |
| SciCode | 38.9% | — |
| WeirdML | 39.5% | — |

## Agentic & Tool Use

- DeepSeek-V3.2-Exp: 37.9 (#31)
- Step 3.5 Flash: —

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| Vending-Bench 2 | 1,034 | — |

## Reasoning

- DeepSeek-V3.2-Exp: 22.0 (#185)
- Step 3.5 Flash: 21.2 (#202)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| NYT Connections (extended) | 36.7% | 28.4% |
| LMArena Hard Prompts | 1433 | 1411 |
| ARC-AGI-2 | 4% | — |
| Kagi LLM Benchmark | 52.2% | — |
| ARC-AGI-1 | 57% | — |
| CritPt | 2.9% | — |
| Chess Puzzles | 14% | — |
| Thematic Generalization | 65% | — |
| DTBench | 87.7% | — |
| LMCA | 29.1% | — |
| Epoch Capabilities Index | 146.28 | — |

## Math

- DeepSeek-V3.2-Exp: 40.8 (#75)
- Step 3.5 Flash: 40.6 (#77)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| MathArena Final-Answer Competitions | 57.7% | 66.8% |
| LMArena Math | 1435 | 1408 |
| OTIS Mock AIME 2024-2025 | 87.8% | — |
| ProofBench | 8% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |

## Knowledge

- DeepSeek-V3.2-Exp: 51.6 (#61)
- Step 3.5 Flash: 39.0 (#127)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| LMArena Expert | 1436 | 1422 |
| GPQA Diamond | 83.4% | — |
| Vectara Hallucination Rate | 5.3% | — |

## Multilingual

- DeepSeek-V3.2-Exp: 52.4 (#85)
- Step 3.5 Flash: 50.7 (#115)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| LMArena Non-English | 1409 | 1386 |
| LMArena Chinese | 1461 | 1448 |
| LMArena French | 1435 | 1421 |
| LMArena German | 1440 | 1404 |
| LMArena Japanese | 1377 | 1353 |
| LMArena Korean | 1372 | 1350 |
| LMArena Russian | 1423 | 1385 |
| LMArena Spanish | 1441 | 1419 |

## Instruction Following

- DeepSeek-V3.2-Exp: 74.7 (#89)
- Step 3.5 Flash: 73.3 (#119)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| LMArena Instruction Following | 1413 | 1385 |

## Long Context

- DeepSeek-V3.2-Exp: 47.6 (#15)
- Step 3.5 Flash: 42.8 (#112)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| LMArena Longer Query | 1428 | 1402 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |

## Writing & Preference

- DeepSeek-V3.2-Exp: 62.5 (#73)
- Step 3.5 Flash: 59.0 (#108)

| Benchmark | DeepSeek-V3.2-Exp | Step 3.5 Flash |
|---|---|---|
| LMArena Text | 1425 | 1403 |
| LMArena Creative Writing | 1403 | 1357 |
| LMArena Multi-Turn | 1427 | 1405 |
| EQ-Bench Creative Writing | 1515 | — |

## FAQ

### Is DeepSeek-V3.2-Exp better than Step 3.5 Flash?

DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 41.7 on the Noometry Index. Step 3.5 Flash costs 1.9× less per token, which makes it the better buy when DeepSeek-V3.2-Exp's lead doesn't matter for your workload.

### Which is cheaper, DeepSeek-V3.2-Exp 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; DeepSeek-V3.2-Exp lists at $0.26 and $0.38.

### Is DeepSeek-V3.2-Exp or Step 3.5 Flash better for coding?

DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.4 versus 41.7 in the Noometry coding category.

### Which has the bigger context window?

Step 3.5 Flash does, with 256K tokens against 164K.

### How many benchmarks do DeepSeek-V3.2-Exp and Step 3.5 Flash share?

19 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and Step 3.5 Flash has 19.