Step 5 Preview vs Hy4
Hy4 leads Step 5 Preview on coding, 65.2 to 61.2 on the Vibe Coding Index, and Hy4 costs 2.2x less per output token ($1.2505 vs $2.70 per million). Updated November 5, 2026; 6 benchmarks have a result for both.
Step 5 Preview or Hy4: which to pick
Pick Hy4 for coding: it scores 65.2 on the Vibe Coding Index against 61.2 for Step 5 Preview, though on 2 benchmarks against 4.
Pick Step 5 Preview for work measured by Harvey's Legal Agent Benchmark (Vals AI), where it leads.
- Shared results
- 6
- Step 5 Preview only
- 61
- Hy4 only
- 0
- Like-for-like categories
- 3 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
Hy4
Hy4 scores 65.2 against 61.2.
4 and 2 benchmarks behind the scores
Intelligence
AA Intelligence Index
No call
Only Step 5 Preview has a AA Intelligence Index result, 28.5.
Directional only
General
Vals Index
Hy4
Hy4 leads, 59.4% against 54.8%.
Same board, both measured
Long documents
Listed context window
No call
Both take 1.05M tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Hy4
Step 5 Preview $0.0024, Hy4 $0.0010 per turn.
List prices
At a glance
- Coding index
- Step 5 Preview 61.2 from 4 benchmarks, Hy4 65.2 from 2 benchmarks.
- Widest category
- Coding, 10.4 points apart on average over 2 shared benchmarks.
- Widest single gap
- Vibe Code Bench (Vals AI): Step 5 Preview 70.4% against Hy4 81.0%, 10.6 points apart on a 0 to 100 scale.
- Output price
- Step 5 Preview $2.70 against Hy4 $1.2505 per million output tokens, 2.2x 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 | Step 5 Preview | Bars | Hy4 | Ties |
|---|---|---|---|---|
| Coding | 0 | 2 | ||
| Agentic work | 1 | 2 | ||
| General | 0 | 1 | ||
| All shared boards | 1 | 5 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index02
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| LiveBench coding | 85.6% | Not measured | Only Step 5 Preview measured |
| LMArena WebDev | 1571 | Not measured | Only Step 5 Preview measured |
| Terminal-Bench 4 (Vals AI) | 31.8% | 39.4% | Hy4 |
| Vibe Code Bench (Vals AI) | 70.4% | 81.0% | Hy4 |
| Code migration (Vals AI) | 39.0% | Not measured | Only Step 5 Preview measured |
| IOI (Vals AI) | 38.1% | Not measured | Only Step 5 Preview measured |
| SWE-bench Multimodal (Vals AI) | 40.8% | Not measured | Only Step 5 Preview measured |
| SciCode | 58.9% | Not measured | Only Step 5 Preview measured |
| LMArena Text: coding | 1482 | Not measured | Only Step 5 Preview measured |
| LMArena Text: software and IT | 1472 | Not measured | Only Step 5 Preview measured |
| LMArena WebDev: React | 1586 | Not measured | Only Step 5 Preview measured |
| LMArena WebDev: HTML | 1555 | Not measured | Only Step 5 Preview measured |
| LMArena WebDev: from an image | 1550 | Not measured | Only Step 5 Preview measured |
| GameBench 2 | 52.2 | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 36 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| AA Intelligence Index | 28.5 | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 1 board here | |||
Agentic workLike-for-likePublic boards, shown as published, kept out of the index12
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| LiveBench research | 72.8% | Not measured | Only Step 5 Preview measured |
| LiveBench browser | 24.8% | Not measured | Only Step 5 Preview measured |
| LiveBench long agentic | 19.4% | Not measured | Only Step 5 Preview measured |
| SREBench (Vals AI) | 0.4% | Not measured | Only Step 5 Preview measured |
| Finance Agent (Vals AI) | 50.7% | Not measured | Only Step 5 Preview measured |
| Tax Agent Bench (Vals AI) | 27.3% | 34.4% | Hy4 |
| Legal Research Bench (Vals AI) | 40.4% | 41.4% | Hy4 |
| Harvey's Legal Agent Benchmark (Vals AI) | 12.9% | 12.5% | Step 5 Preview |
| Excel Modeling Benchmark (Vals AI) | 57.1% | Not measured | Only Step 5 Preview measured |
| BioMysteryBench (Vals AI) | 67.4% | Not measured | Only Step 5 Preview measured |
| Terminal-Bench-Science (Vals AI) | 2.9% | Not measured | Only Step 5 Preview measured |
| Vals Science Index | 29.6% | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 32 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| LiveBench reasoning | 83.8% | Not measured | Only Step 5 Preview measured |
| MysteryMechanism (Vals AI) | 18.5% | Not measured | Only Step 5 Preview measured |
| Humanity's Last Exam (Artificial Analysis run) | 46.5% | Not measured | Only Step 5 Preview measured |
| LMArena Text: hard prompts | 1463 | Not measured | Only Step 5 Preview measured |
| LMArena Text: expert | 1478 | Not measured | Only Step 5 Preview measured |
| LMArena Text: hard prompts in English | 1461 | Not measured | Only Step 5 Preview measured |
| LMArena Vision: diagrams | 1308 | Not measured | Only Step 5 Preview measured |
| LMArena Vision: homework | 1265 | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 16 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index3 boards, none shared
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| ProofBench | 42.0% | Not measured | Only Step 5 Preview measured |
| LMArena Text: math | 1457 | Not measured | Only Step 5 Preview measured |
| LMArena Text: mathematical work | 1462 | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 11 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index4 boards, none shared
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| LMArena Text: science | 1459 | Not measured | Only Step 5 Preview measured |
| LMArena Text: medicine and healthcare | 1439 | Not measured | Only Step 5 Preview measured |
| LMArena Text: legal and government | 1441 | Not measured | Only Step 5 Preview measured |
| LMArena Text: business and finance | 1443 | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 16 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| LiveBench design | 76.9% | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 8 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Step 5 Preview | Hy4 | Leader |
|---|---|---|---|
| Vals Index | 54.8% | 59.4% | Hy4 |
| LiveBench data analysis | 63.5% | Not measured | Only Step 5 Preview measured |
| Public Benefits Bench (Vals AI) | 58.1% | Not measured | Only Step 5 Preview measured |
| Long context reasoning (AA-LCR) | 88.3% | Not measured | Only Step 5 Preview measured |
| LMArena Text | 1443 | Not measured | Only Step 5 Preview measured |
| LMArena Text: instruction following | 1441 | Not measured | Only Step 5 Preview measured |
| LMArena Text: creative writing | 1413 | Not measured | Only Step 5 Preview measured |
| LMArena Vision | 1265 | Not measured | Only Step 5 Preview measured |
| LMArena Text: English | 1454 | Not measured | Only Step 5 Preview measured |
| LMArena Text: non-English | 1429 | Not measured | Only Step 5 Preview measured |
| LMArena Text: Chinese | 1502 | Not measured | Only Step 5 Preview measured |
| LMArena Text: German | 1454 | Not measured | Only Step 5 Preview measured |
| LMArena Text: Japanese | 1398 | Not measured | Only Step 5 Preview measured |
| LMArena Text: Korean | 1415 | Not measured | Only Step 5 Preview measured |
| LMArena Text: French | 1440 | Not measured | Only Step 5 Preview measured |
| LMArena Text: Spanish | 1444 | Not measured | Only Step 5 Preview measured |
| LMArena Text: Russian | 1431 | Not measured | Only Step 5 Preview measured |
| LMArena Text: Polish | 1390 | Not measured | Only Step 5 Preview measured |
| LMArena Text: multi-turn | 1437 | Not measured | Only Step 5 Preview measured |
| LMArena Text: longer queries | 1452 | Not measured | Only Step 5 Preview measured |
| LMArena Text: writing and language | 1427 | Not measured | Only Step 5 Preview measured |
| LMArena Text: entertainment and media | 1412 | Not measured | Only Step 5 Preview measured |
| LMArena Vision: English | 1290 | Not measured | Only Step 5 Preview measured |
| LMArena Vision: reading text | 1282 | Not measured | Only Step 5 Preview measured |
| Neither model has a result on the other 24 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Step 5 Preview | Hy4 | 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
Hy4 costs less
Repository review
50K input + 3K output tokens
Hy4 costs less
Agent loop
200K input + 10K output tokens
Hy4 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 | StepFun Step 5 Preview | Tencent Hy4 |
|---|---|---|
| Maker | StepFun | Tencent |
| Weights | Open weights | Open weights |
| Released | October 8, 2026 | November 4, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $1.00 | $0.417 |
| Output $/Mtok | $2.70 | $1.2505 |
| Coding index | 61.2 | 65.2 |
| Independent measurements behind it | 4 | 2 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Step 5 Preview better than Hy4 for coding?
Pick Hy4 for coding: it scores 65.2 on the Vibe Coding Index against 61.2 for Step 5 Preview, though on 2 benchmarks against 4. Pick Step 5 Preview for work measured by Harvey's Legal Agent Benchmark (Vals AI), where it leads.
Which is cheaper, Step 5 Preview or Hy4?
Step 5 Preview costs $1.00 per million input tokens and $2.70 per million output tokens; Hy4 costs $0.417 and $1.2505. Cost per solved task is known for neither of them.
Which has the bigger context window, Step 5 Preview or Hy4?
Neither: both take 1.05M tokens.
How many benchmarks is the Step 5 Preview vs Hy4 comparison based on?
6 benchmarks have a result for both models as of November 5, 2026. Step 5 Preview's coding index rests on 4 independent benchmarks and Hy4's on 2. Where only one model has a result, the page says so instead of naming a winner.