Model comparison
DeepSeek V4.1 Flash vs Step 5 Preview
DeepSeek V4.1 Flash is the stronger model overall, scoring 52.8 to 47.9 on the Noometry Index.
Last verified . 18 shared benchmarks.
DeepSeek
StepFunSummary
- They share 18 benchmarks with published results for both. DeepSeek V4.1 Flash scores higher in 8 categories and Step 5 Preview in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4.1 Flash leads 66.7 to 46.1.
- The biggest single-benchmark swing is ProofBench: 54% for DeepSeek V4.1 Flash and 42% for Step 5 Preview.
- DeepSeek V4.1 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $1 / $2.70 for Step 5 Preview.
- Step 5 Preview accepts more context: 1.02M tokens versus 1M.
- DeepSeek V4.1 Flash has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4.1 Flash | Step 5 Preview | |
|---|---|---|
| Provider | DeepSeek | StepFun |
| Noometry Index | 52.8 | 47.9 |
| Released | 2026-09-09 | 2026-09-16 |
| Weights | Open | Proprietary |
| Context window | 1M | 1.02M |
| Max output | 393K | 66K |
| Input $ / M tokens | $0.15 | $1 |
| Output $ / M tokens | $0.60 | $2.70 |
| Results tracked | 37 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 52.9 (#32), Step 5 Preview: 51.8 (#36)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| LMArena WebDev | 1619 | 1564 |
| SciCode | 51.9% | 58.9% |
| LMArena Coding | 1506 | 1480 |
| ALE-Bench | 1,092 | — |
Agentic & Tool Use Not comparable
DeepSeek V4.1 Flash: 31.2 (#69), Step 5 Preview: —
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| APEX-Agents | 39.5% | — |
| GDP.pdf | 19.8% | — |
Reasoning DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 50.2 (#36), Step 5 Preview: 40.0 (#57)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| CritPt | 14.3% | 20.9% |
| LMArena Hard Prompts | 1483 | 1465 |
| NYT Connections (extended) | 89.6% | — |
| Mystery Game Puzzles | 43% | — |
| DTBench | 89.9% | — |
| LMCA | 47% | — |
| Surface Evolver Bench | 46.3% | — |
| Epoch Capabilities Index | 154.9 | — |
Math DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 66.7 (#25), Step 5 Preview: 46.1 (#72)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| ProofBench | 54% | 42% |
| LMArena Math | 1477 | 1470 |
| FrontierMath (Tiers 1-3) | 67.4% | — |
| FrontierMath Tier 4 | 26.8% | — |
| OTIS Mock AIME 2024-2025 | 98.3% | — |
Knowledge DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 57.9 (#38), Step 5 Preview: 41.2 (#112)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| LMArena Expert | 1506 | 1470 |
| GPQA Diamond | 89.8% | — |
Multimodal Step 5 Preview leads
DeepSeek V4.1 Flash: 39.1 (#61), Step 5 Preview: 41.0 (#41)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| LMArena Vision | 1277 | 1267 |
| Furniture Assembly | 34.2% | — |
Multilingual DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 55.0 (#35), Step 5 Preview: 53.8 (#54)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| LMArena Non-English | 1448 | 1432 |
| LMArena Chinese | 1497 | 1519 |
| LMArena Russian | 1471 | 1436 |
| LMArena Spanish | 1459 | 1439 |
| LMArena French | 1452 | — |
| LMArena German | 1484 | — |
| LMArena Japanese | 1412 | — |
| LMArena Korean | 1452 | — |
Instruction Following DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 77.3 (#26), Step 5 Preview: 76.0 (#50)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| LMArena Instruction Following | 1474 | 1444 |
Long Context Too close to call
DeepSeek V4.1 Flash: 45.2 (#47), Step 5 Preview: 44.5 (#64)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| LMArena Longer Query | 1475 | 1455 |
Writing & Preference DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 65.4 (#48), Step 5 Preview: 62.8 (#71)
| Benchmark | DeepSeek V4.1 Flash | Step 5 Preview |
|---|---|---|
| LMArena Text | 1462 | 1442 |
| LMArena Creative Writing | 1435 | 1410 |
| LMArena Multi-Turn | 1457 | 1447 |
| EQ-Bench Creative Writing | 1540 | — |
Frequently asked questions
Is DeepSeek V4.1 Flash better than Step 5 Preview?
DeepSeek V4.1 Flash is the stronger model overall, scoring 52.8 to 47.9 on the Noometry Index.
Which is cheaper, DeepSeek V4.1 Flash or Step 5 Preview?
DeepSeek V4.1 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; Step 5 Preview lists at $1 and $2.70.
Is DeepSeek V4.1 Flash or Step 5 Preview better for coding?
DeepSeek V4.1 Flash scores higher on coding benchmarks: 52.9 versus 51.8 in the Noometry coding category.
Which has the bigger context window?
Step 5 Preview does, with 1.02M tokens against 1M.
How many benchmarks do DeepSeek V4.1 Flash and Step 5 Preview share?
18 benchmarks have published results for both models. DeepSeek V4.1 Flash has 37 scored results on Noometry and Step 5 Preview has 18.