Model comparison
DeepSeek-V3.2-Speciale vs Step 5 Preview
Step 5 Preview is the stronger model overall, scoring 47.9 to 39.7 on the Noometry Index. DeepSeek-V3.2-Speciale costs 1.7× less per token, which makes it the better buy when Step 5 Preview's lead doesn't matter for your workload.
Last verified . 0 shared benchmarks.
DeepSeek
StepFunSummary
- The widest gap is in writing & preference, where Step 5 Preview leads 62.8 to 46.0.
- DeepSeek-V3.2-Speciale is cheaper at $0.58 / $1.68 per million input/output tokens, against $1 / $2.70 for Step 5 Preview.
- Step 5 Preview accepts more context: 1.02M tokens versus 128K.
- DeepSeek-V3.2-Speciale has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Speciale | Step 5 Preview | |
|---|---|---|
| Provider | DeepSeek | StepFun |
| Noometry Index | 39.7 | 47.9 |
| Released | 2025-12-01 | 2026-09-16 |
| Weights | Open | Proprietary |
| Context window | 128K | 1.02M |
| Max output | 128K | 66K |
| Input $ / M tokens | $0.58 | $1 |
| Output $ / M tokens | $1.68 | $2.70 |
| Results tracked | 3 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Step 5 Preview leads
DeepSeek-V3.2-Speciale: 40.4 (#140), Step 5 Preview: 51.8 (#36)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| LMArena WebDev | — | 1564 |
| SciCode | — | 58.9% |
| WeirdML | 46.7% | — |
| LMArena Coding | — | 1480 |
Reasoning Step 5 Preview leads
DeepSeek-V3.2-Speciale: 32.9 (#73), Step 5 Preview: 40.0 (#57)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| SimpleBench | 52.6% | — |
| CritPt | — | 20.9% |
| LMArena Hard Prompts | — | 1465 |
Math Not comparable
DeepSeek-V3.2-Speciale: —, Step 5 Preview: 46.1 (#72)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| ProofBench | — | 42% |
| LMArena Math | — | 1470 |
Knowledge Not comparable
DeepSeek-V3.2-Speciale: —, Step 5 Preview: 41.2 (#112)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| LMArena Expert | — | 1470 |
Multimodal Not comparable
DeepSeek-V3.2-Speciale: —, Step 5 Preview: 41.0 (#41)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| LMArena Vision | — | 1267 |
Multilingual Not comparable
DeepSeek-V3.2-Speciale: —, Step 5 Preview: 53.8 (#54)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| LMArena Non-English | — | 1432 |
| LMArena Chinese | — | 1519 |
| LMArena Russian | — | 1436 |
| LMArena Spanish | — | 1439 |
Instruction Following Not comparable
DeepSeek-V3.2-Speciale: —, Step 5 Preview: 76.0 (#50)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| LMArena Instruction Following | — | 1444 |
Long Context Not comparable
DeepSeek-V3.2-Speciale: —, Step 5 Preview: 44.5 (#64)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| LMArena Longer Query | — | 1455 |
Writing & Preference Step 5 Preview leads
DeepSeek-V3.2-Speciale: 46.0 (#222), Step 5 Preview: 62.8 (#71)
| Benchmark | DeepSeek-V3.2-Speciale | Step 5 Preview |
|---|---|---|
| LMArena Text | — | 1442 |
| LMArena Creative Writing | — | 1410 |
| EQ-Bench Creative Writing | 1276 | — |
| LMArena Multi-Turn | — | 1447 |
Frequently asked questions
Is DeepSeek-V3.2-Speciale better than Step 5 Preview?
Step 5 Preview is the stronger model overall, scoring 47.9 to 39.7 on the Noometry Index. DeepSeek-V3.2-Speciale costs 1.7× less per token, which makes it the better buy when Step 5 Preview's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Speciale or Step 5 Preview?
DeepSeek-V3.2-Speciale is cheaper. It lists at $0.58 per million input tokens and $1.68 per million output tokens; Step 5 Preview lists at $1 and $2.70.
Is DeepSeek-V3.2-Speciale or Step 5 Preview better for coding?
Step 5 Preview scores higher on coding benchmarks: 51.8 versus 40.4 in the Noometry coding category.
Which has the bigger context window?
Step 5 Preview does, with 1.02M tokens against 128K.
How many benchmarks do DeepSeek-V3.2-Speciale and Step 5 Preview share?
0 benchmarks have published results for both models. DeepSeek-V3.2-Speciale has 3 scored results on Noometry and Step 5 Preview has 18.