DeepSeek V4 Flash 0731 vs Step 5 Preview
Step 5 Preview leads DeepSeek V4 Flash 0731 on coding, 61.1 to 51.8 on the Vibe Coding Index, and DeepSeek V4 Flash 0731 costs 9.6x less per output token ($0.28 vs $2.70 per million). Updated October 21, 2026; 15 benchmarks have a result for both.
DeepSeek V4 Flash 0731 or Step 5 Preview: which to pick
Pick Step 5 Preview for coding: it scores 61.1 on the Vibe Coding Index against 51.8 for DeepSeek V4 Flash 0731.
Pick DeepSeek V4 Flash 0731 for output tokens 9.6x cheaper ($0.28 vs $2.70 per million) and work measured by Vibe Code Bench (Vals AI) or Tax Agent Bench (Vals AI), where it leads.
- Shared results
- 15
- DeepSeek V4 Flash 0731 only
- 40
- Step 5 Preview only
- 46
- Like-for-like categories
- 6 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
Step 5 Preview
Step 5 Preview scores 61.1 against 51.8.
4 benchmarks behind each score
Intelligence
AA Intelligence Index
Step 5 Preview
Step 5 Preview leads, 43.7 against 34.3.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only DeepSeek V4 Flash 0731 has a AA Agentic Index result, 41.0.
Directional only
Reasoning
GPQA Diamond
No call
Only DeepSeek V4 Flash 0731 has a GPQA Diamond result, 89.9%.
Directional only
Knowledge
MMLU Pro
No call
Only DeepSeek V4 Flash 0731 has a MMLU Pro result, 86.2%.
Directional only
Web design
Design Arena: websites
No call
Only DeepSeek V4 Flash 0731 has a Design Arena: websites result, 1238.
Directional only
Long documents
Listed context window
No call
Both take 1.05M tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V4 Flash 0731
DeepSeek V4 Flash 0731 $0.0003, Step 5 Preview $0.0024 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only DeepSeek V4 Flash 0731 has the number, $0.01.
Directional only
At a glance
- Coding index
- DeepSeek V4 Flash 0731 51.8 from 4 benchmarks, Step 5 Preview 61.1 from 4 benchmarks.
- Widest category
- Math, 14.0 points apart on average over 1 shared benchmark.
- Widest single gap
- Terminal-Bench 4 (Vals AI): DeepSeek V4 Flash 0731 18.7% against Step 5 Preview 31.8%, 17.5 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V4 Flash 0731 $0.28 against Step 5 Preview $2.70 per million output tokens, 9.6x apart.
- Context window
- Both take 1.05M tokens.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | DeepSeek V4 Flash 0731 | Bars | Step 5 Preview | Ties |
|---|---|---|---|---|
| Coding | 1 | 4 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 2 | 4 | ||
| Reasoning | 0 | 1 | ||
| Math | 1 | 0 | ||
| General | 0 | 1 | ||
| All shared boards | 4 | 11 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index14
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 85.6% | Only Step 5 Preview measured |
| LMArena WebDev | Not measured | 1567 | Only Step 5 Preview measured |
| AA Coding Index | 69.1 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LiveCodeBench (Vals AI) | 87.3% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| SWE-bench Verified (Vals AI) | 88.8% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Terminal-Bench 4 (Vals AI) | 18.7% | 31.8% | Step 5 Preview |
| Vibe Code Bench (Vals AI) | 74.7% | 70.4% | DeepSeek V4 Flash 0731 |
| Code migration (Vals AI) | 38.6% | 39.0% | Step 5 Preview |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| IOI (Vals AI) | 32.7% | 38.1% | Step 5 Preview |
| Terminal-Bench 2.1 (Vals AI) | 67.0% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| SWE-bench Multimodal (Vals AI) | Not measured | 40.8% | Only Step 5 Preview measured |
| SkillsBench (Vals AI) | 50.7% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| CyberBench (Vals AI) | 72.1% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| SciCode | 50.3% | 58.9% | Step 5 Preview |
| FrontierCode | 18.8% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| WeirdML | 63.0% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| ALE-Bench | 1306.1 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LMArena Text: coding | Not measured | 1485 | Only Step 5 Preview measured |
| LMArena Text: software and IT | Not measured | 1473 | Only Step 5 Preview measured |
| LMArena WebDev: React | Not measured | 1583 | Only Step 5 Preview measured |
| LMArena WebDev: HTML | Not measured | 1556 | Only Step 5 Preview measured |
| LMArena WebDev: from an image | Not measured | 1550 | Only Step 5 Preview measured |
| Neither model has a result on the other 15 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| AA Intelligence Index | 34.3 | 43.7 | Step 5 Preview |
| Epoch Capabilities Index | 154.8 | Not measured | Only DeepSeek V4 Flash 0731 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index24
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| AA Agentic Index | 41.0 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LiveBench research | Not measured | 72.8% | Only Step 5 Preview measured |
| LiveBench browser | Not measured | 24.0% | Only Step 5 Preview measured |
| LiveBench long agentic | Not measured | 27.5% | Only Step 5 Preview measured |
| SREBench (Vals AI) | Not measured | 0.4% | Only Step 5 Preview measured |
| Finance Agent (Vals AI) | 49.5% | 50.7% | Step 5 Preview |
| Tax Agent Bench (Vals AI) | 28.1% | 27.3% | DeepSeek V4 Flash 0731 |
| Legal Research Bench (Vals AI) | 30.3% | 40.4% | Step 5 Preview |
| Excel Modeling Benchmark (Vals AI) | 57.0% | 57.1% | Step 5 Preview |
| BioMysteryBench (Vals AI) | 64.4% | 67.4% | Step 5 Preview |
| Terminal-Bench-Science (Vals AI) | 4.3% | 2.9% | DeepSeek V4 Flash 0731 |
| Vals Science Index | Not measured | 29.6% | Only Step 5 Preview measured |
| Neither model has a result on the other 27 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| GPQA Diamond | 89.9% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LiveBench reasoning | Not measured | 83.8% | Only Step 5 Preview measured |
| MysteryMechanism (Vals AI) | Not measured | 18.5% | Only Step 5 Preview measured |
| GPQA Diamond (Artificial Analysis run) | 90.8% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Humanity's Last Exam (Artificial Analysis run) | 38.6% | 46.5% | Step 5 Preview |
| GPQA Diamond (Epoch AI run) | 91.0% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| ARC-AGI | 89.0% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| ARC-AGI-2 | 61.4% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| SimpleBench | 61.1% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Chess puzzles | 32.0% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Mystery game puzzles | 34.0% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LMArena Text: hard prompts | Not measured | 1461 | Only Step 5 Preview measured |
| LMArena Text: expert | Not measured | 1485 | Only Step 5 Preview measured |
| LMArena Text: hard prompts in English | Not measured | 1454 | Only Step 5 Preview measured |
| LMArena Vision: diagrams | Not measured | 1308 | Only Step 5 Preview measured |
| LMArena Vision: homework | Not measured | 1265 | Only Step 5 Preview measured |
| Neither model has a result on the other 5 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| OTIS Mock AIME | 94.4% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| FrontierMath, tiers 1 to 3 | 57.5% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| FrontierMath, tier 4 | 24.4% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| ProofBench | 56.0% | 42.0% | DeepSeek V4 Flash 0731 |
| LMArena Text: math | Not measured | 1458 | Only Step 5 Preview measured |
| LMArena Text: mathematical work | Not measured | 1460 | Only Step 5 Preview measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index10 boards, none shared
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| MMLU Pro | 86.2% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LegalBench (Vals AI) | 77.7% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| CorpFin (Vals AI) | 61.9% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| TaxEval (Vals AI) | 70.7% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| MedCode (Vals AI) | 41.4% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| SimpleQA Verified | 33.6% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LMArena Text: science | Not measured | 1454 | Only Step 5 Preview measured |
| LMArena Text: medicine and healthcare | Not measured | 1439 | Only Step 5 Preview measured |
| LMArena Text: legal and government | Not measured | 1437 | Only Step 5 Preview measured |
| LMArena Text: business and finance | Not measured | 1435 | Only Step 5 Preview measured |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| Design Arena: websites | 1238 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Design Arena: UI components | 1235 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Design Arena: game dev | 1208 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Design Arena: data viz | 1180 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Design Arena: 3D | 1209 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Design Arena: all code categories | 1230 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Design Arena: SVG | 1185 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Design Arena: ASCII art | 1090 | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LiveBench design | Not measured | 76.9% | Only Step 5 Preview measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| LiveBench data analysis | Not measured | 63.5% | Only Step 5 Preview measured |
| Public Benefits Bench (Vals AI) | Not measured | 58.1% | Only Step 5 Preview measured |
| MedScribe (Vals AI) | 80.4% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| Long context reasoning (AA-LCR) | 79.7% | 88.3% | Step 5 Preview |
| DTBench | 90.9% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LMCA | 41.7% | Not measured | Only DeepSeek V4 Flash 0731 measured |
| LMArena Text | Not measured | 1440 | Only Step 5 Preview measured |
| LMArena Text: instruction following | Not measured | 1438 | Only Step 5 Preview measured |
| LMArena Text: creative writing | Not measured | 1409 | Only Step 5 Preview measured |
| LMArena Vision | Not measured | 1265 | Only Step 5 Preview measured |
| LMArena Text: English | Not measured | 1447 | Only Step 5 Preview measured |
| LMArena Text: non-English | Not measured | 1428 | Only Step 5 Preview measured |
| LMArena Text: Chinese | Not measured | 1500 | Only Step 5 Preview measured |
| LMArena Text: Japanese | Not measured | 1383 | Only Step 5 Preview measured |
| LMArena Text: French | Not measured | 1438 | Only Step 5 Preview measured |
| LMArena Text: Spanish | Not measured | 1441 | Only Step 5 Preview measured |
| LMArena Text: Russian | Not measured | 1433 | Only Step 5 Preview measured |
| LMArena Text: multi-turn | Not measured | 1437 | Only Step 5 Preview measured |
| LMArena Text: longer queries | Not measured | 1449 | Only Step 5 Preview measured |
| LMArena Text: writing and language | Not measured | 1425 | Only Step 5 Preview measured |
| LMArena Text: entertainment and media | Not measured | 1412 | Only Step 5 Preview measured |
| LMArena Vision: English | Not measured | 1290 | Only Step 5 Preview measured |
| LMArena Vision: reading text | Not measured | 1282 | Only Step 5 Preview measured |
| Neither model has a result on the other 23 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V4 Flash 0731 | Step 5 Preview | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
DeepSeek V4 Flash 0731 costs less
Repository review
50K input + 3K output tokens
DeepSeek V4 Flash 0731 costs less
Agent loop
200K input + 10K output tokens
DeepSeek V4 Flash 0731 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | DeepSeek V4 Flash 0731 | StepFun Step 5 Preview |
|---|---|---|
| Maker | DeepSeek | StepFun |
| Weights | Open weights | Open weights |
| Released | July 31, 2026 | October 8, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $0.14 | $1.00 |
| Output $/Mtok | $0.28 | $2.70 |
| Coding index | 51.8 | 61.1 |
| Independent measurements behind it | 4 | 4 |
| Cost per solved task | $0.01 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V4 Flash 0731 better than Step 5 Preview for coding?
Pick Step 5 Preview for coding: it scores 61.1 on the Vibe Coding Index against 51.8 for DeepSeek V4 Flash 0731. Pick DeepSeek V4 Flash 0731 for output tokens 9.6x cheaper ($0.28 vs $2.70 per million) and work measured by Vibe Code Bench (Vals AI) or Tax Agent Bench (Vals AI), where it leads.
Which is cheaper, DeepSeek V4 Flash 0731 or Step 5 Preview?
DeepSeek V4 Flash 0731 costs $0.14 per million input tokens and $0.28 per million output tokens; Step 5 Preview costs $1.00 and $2.70. Cost per solved task is known for DeepSeek V4 Flash 0731 only.
Which has the bigger context window, DeepSeek V4 Flash 0731 or Step 5 Preview?
Neither: both take 1.05M tokens.
How many benchmarks is the DeepSeek V4 Flash 0731 vs Step 5 Preview comparison based on?
15 benchmarks have a result for both models as of October 21, 2026. DeepSeek V4 Flash 0731's coding index rests on 4 independent benchmarks and Step 5 Preview's on 4. Where only one model has a result, the page says so instead of naming a winner.