Model comparison
DeepSeek V4 Flash vs Step 5 Preview
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 47.9 on the Noometry Index.
Last verified . 17 shared benchmarks.
DeepSeek
StepFunSummary
- They share 17 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 4 categories and Step 5 Preview in 4 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek V4 Flash leads 55.4 to 41.2.
- The biggest single-benchmark swing is ProofBench: 56% for DeepSeek V4 Flash and 42% for Step 5 Preview.
- DeepSeek V4 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 Flash has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Flash | Step 5 Preview | |
|---|---|---|
| Provider | DeepSeek | StepFun |
| Noometry Index | 53.6 | 47.9 |
| Released | 2026-04-24 | 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 | 41 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Step 5 Preview leads
DeepSeek V4 Flash: 47.9 (#59), Step 5 Preview: 51.8 (#36)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| LMArena WebDev | 1582 | 1564 |
| SciCode | 49.9% | 58.9% |
| LMArena Coding | 1457 | 1480 |
| FrontierCode | 18.8% | — |
| WeirdML | 63% | — |
| ALE-Bench | 1,306 | — |
Reasoning DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.7 (#30), Step 5 Preview: 40.0 (#57)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| CritPt | 16.6% | 20.9% |
| LMArena Hard Prompts | 1444 | 1465 |
| ARC-AGI-2 | 61.4% | — |
| SimpleBench | 61.1% | — |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 89.6% | — |
| ARC-AGI-1 | 89% | — |
| Chess Puzzles | 33% | — |
| Mystery Game Puzzles | 34% | — |
| DTBench | 90.9% | — |
| LMCA | 41.7% | — |
| Epoch Capabilities Index | 154.49 | — |
Math DeepSeek V4 Flash leads
DeepSeek V4 Flash: 60.3 (#37), Step 5 Preview: 46.1 (#72)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| ProofBench | 56% | 42% |
| LMArena Math | 1427 | 1470 |
| FrontierMath (Tiers 1-3) | 57.5% | — |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
| OTIS Mock AIME 2024-2025 | 94.4% | — |
Knowledge DeepSeek V4 Flash leads
DeepSeek V4 Flash: 55.4 (#48), Step 5 Preview: 41.2 (#112)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| LMArena Expert | 1441 | 1470 |
| GPQA Diamond | 91% | — |
| SimpleQA Verified | 33.6% | — |
Multimodal Not comparable
DeepSeek V4 Flash: —, Step 5 Preview: 41.0 (#41)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| LMArena Vision | — | 1267 |
Multilingual Too close to call
DeepSeek V4 Flash: 53.0 (#72), Step 5 Preview: 53.8 (#54)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| LMArena Non-English | 1420 | 1432 |
| LMArena Chinese | 1468 | 1519 |
| LMArena Russian | 1428 | 1436 |
| LMArena Spanish | 1436 | 1439 |
| LMArena French | 1439 | — |
| LMArena German | 1418 | — |
| LMArena Japanese | 1406 | — |
| LMArena Korean | 1384 | — |
Instruction Following Step 5 Preview leads
DeepSeek V4 Flash: 74.9 (#81), Step 5 Preview: 76.0 (#50)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| LMArena Instruction Following | 1421 | 1444 |
Long Context Too close to call
DeepSeek V4 Flash: 43.8 (#85), Step 5 Preview: 44.5 (#64)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| LMArena Longer Query | 1434 | 1455 |
Writing & Preference DeepSeek V4 Flash leads
DeepSeek V4 Flash: 63.8 (#61), Step 5 Preview: 62.8 (#71)
| Benchmark | DeepSeek V4 Flash | Step 5 Preview |
|---|---|---|
| LMArena Text | 1432 | 1442 |
| LMArena Creative Writing | 1403 | 1410 |
| LMArena Multi-Turn | 1449 | 1447 |
| EQ-Bench Creative Writing | 1559 | — |
Frequently asked questions
Is DeepSeek V4 Flash better than Step 5 Preview?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 47.9 on the Noometry Index.
Which is cheaper, DeepSeek V4 Flash or Step 5 Preview?
DeepSeek V4 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 Flash or Step 5 Preview better for coding?
Step 5 Preview scores higher on coding benchmarks: 51.8 versus 47.9 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 Flash and Step 5 Preview share?
17 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and Step 5 Preview has 18.