Enter opens the highlighted result. Esc closes.

Step 5 Preview vs Hy4 flash

Step 5 Preview leads Hy4 flash on coding, 57.3 to 44.5 on the Vibe Coding Index, and Hy4 flash costs 19.2x less per output token ($0.1405 vs $2.70 per million). Updated November 22, 2026; 7 benchmarks have a result for both.

Step 5 PreviewStepFun
Coding
57.3
Overall
55.1
Output $/M
$2.70
6
Hy4 flashTencent
Coding
44.5
Overall
No data
Output $/M
$0.1405
1
vs boards won of 7 shared

Step 5 Preview or Hy4 flash: which to pick

Pick Step 5 Preview for coding: it scores 57.3 on the Vibe Coding Index against 44.5 for Hy4 flash.

Pick Hy4 flash for output tokens 19.2x cheaper ($0.1405 vs $2.70 per million) and a larger context window (1.05M against 1M tokens).

Shared results
7
Step 5 Preview only
69
Hy4 flash only
5
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.

At a glance

Coding index
Step 5 Preview 57.3 from 4 benchmarks, Hy4 flash 44.5 from 2 benchmarks.
Widest category
Coding, 9.8 points apart on average over 3 shared benchmarks.
Widest single gap
Terminal-Bench 4 (Vals AI): Step 5 Preview 31.8% against Hy4 flash 15.2%, 22.2 points apart on a 0 to 100 scale.
Output price
Step 5 Preview $2.70 against Hy4 flash $0.1405 per million output tokens, 19.2x apart.
Context window
Step 5 Preview takes 1M tokens, Hy4 flash 1.05M.

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.

Boards won, where both have a result
CategoryStep 5 PreviewBarsHy4 flashTies
Coding21
Agentic work30
General10
All shared boards61
CodingLike-for-likeSix of these boards feed the Vibe Coding Index21
Coding benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
LiveBench coding67.6%Not measuredOnly Step 5 Preview measured
LMArena WebDev1570Not measuredOnly Step 5 Preview measured
Terminal-Bench 4 (Vals AI)31.8%15.2%Step 5 Preview
Vibe Code Bench (Vals AI)70.4%74.8%Hy4 flash
Code migration (Vals AI)39.0%Not measuredOnly Step 5 Preview measured
IOI (Vals AI)38.1%Not measuredOnly Step 5 Preview measured
SWE-bench Multimodal (Vals AI)40.8%Not measuredOnly Step 5 Preview measured
SciCode58.9%Not measuredOnly Step 5 Preview measured
Terminal-Bench 4.0 (Artificial Analysis run)33.3%Not measuredOnly Step 5 Preview measured
FrontierCode 2.012.0%Not measuredOnly Step 5 Preview measured
LMArena Text: coding1481Not measuredOnly Step 5 Preview measured
LMArena Text: software and IT1472Not measuredOnly Step 5 Preview measured
LMArena WebDev: React1588Not measuredOnly Step 5 Preview measured
LMArena WebDev: HTML1542Not measuredOnly Step 5 Preview measured
LMArena WebDev: from an image1550Not measuredOnly Step 5 Preview measured
GameBench 253.250.5Step 5 Preview
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
Intelligence benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
AA Intelligence Index28.5Not measuredOnly 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 index30
Agentic work benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
LiveBench research75.5%Not measuredOnly Step 5 Preview measured
LiveBench browser24.8%Not measuredOnly Step 5 Preview measured
LiveBench long agentic19.4%Not measuredOnly Step 5 Preview measured
SREBench (Vals AI)0.4%Not measuredOnly Step 5 Preview measured
Finance Agent (Vals AI)50.7%Not measuredOnly Step 5 Preview measured
Tax Agent Bench (Vals AI)27.3%23.3%Step 5 Preview
Legal Research Bench (Vals AI)40.4%38.5%Step 5 Preview
Harvey's Legal Agent Benchmark (Vals AI)12.9%5.4%Step 5 Preview
Excel Modeling Benchmark (Vals AI)57.1%Not measuredOnly Step 5 Preview measured
BioMysteryBench (Vals AI)67.4%Not measuredOnly Step 5 Preview measured
Terminal-Bench-Science (Vals AI)2.9%Not measuredOnly Step 5 Preview measured
Vals Science Index29.6%Not measuredOnly 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 index10 boards, none shared
Reasoning benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
LiveBench reasoning83.8%Not measuredOnly Step 5 Preview measured
MysteryMechanism (Vals AI)18.5%Not measuredOnly Step 5 Preview measured
Humanity's Last Exam (Artificial Analysis run)46.5%Not measuredOnly Step 5 Preview measured
HLE-Diamond (Artificial Analysis run)24.3%Not measuredOnly Step 5 Preview measured
CritPt v2, hard subset (Artificial Analysis run)32.5%Not measuredOnly Step 5 Preview measured
LMArena Text: hard prompts1463Not measuredOnly Step 5 Preview measured
LMArena Text: expert1479Not measuredOnly Step 5 Preview measured
LMArena Text: hard prompts in English1464Not measuredOnly Step 5 Preview measured
LMArena Vision: diagrams1301Not measuredOnly Step 5 Preview measured
LMArena Vision: homework1275Not measuredOnly 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
Math benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
ProofBench42.0%Not measuredOnly Step 5 Preview measured
LMArena Text: math1466Not measuredOnly Step 5 Preview measured
LMArena Text: mathematical work1466Not measuredOnly 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 index5 boards, none shared
Knowledge benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
AA-Omniscience16.4Not measuredOnly Step 5 Preview measured
LMArena Text: science1463Not measuredOnly Step 5 Preview measured
LMArena Text: medicine and healthcare1443Not measuredOnly Step 5 Preview measured
LMArena Text: legal and government1439Not measuredOnly Step 5 Preview measured
LMArena Text: business and finance1445Not measuredOnly 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 index6 boards, none shared
Web design benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
Design Arena: websitesNot measured1227Only Hy4 flash measured
Design Arena: UI componentsNot measured1262Only Hy4 flash measured
Design Arena: game devNot measured1202Only Hy4 flash measured
Design Arena: 3DNot measured1209Only Hy4 flash measured
Design Arena: all code categoriesNot measured1225Only Hy4 flash measured
LiveBench design70.8%Not measuredOnly Step 5 Preview measured
Neither model has a result on the other 3 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index10
General benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
Vals Index54.8%50.8%Step 5 Preview
LiveBench data analysis63.5%Not measuredOnly Step 5 Preview measured
Public Benefits Bench (Vals AI)58.1%Not measuredOnly Step 5 Preview measured
Long context reasoning (AA-LCR)88.3%Not measuredOnly Step 5 Preview measured
GDPval-AA1591Not measuredOnly Step 5 Preview measured
LMArena Text1444Not measuredOnly Step 5 Preview measured
LMArena Text: instruction following1446Not measuredOnly Step 5 Preview measured
LMArena Text: creative writing1416Not measuredOnly Step 5 Preview measured
LMArena Vision1265Not measuredOnly Step 5 Preview measured
LMArena Text: English1456Not measuredOnly Step 5 Preview measured
LMArena Text: non-English1430Not measuredOnly Step 5 Preview measured
LMArena Text: Chinese1495Not measuredOnly Step 5 Preview measured
LMArena Text: German1453Not measuredOnly Step 5 Preview measured
LMArena Text: Japanese1398Not measuredOnly Step 5 Preview measured
LMArena Text: Korean1402Not measuredOnly Step 5 Preview measured
LMArena Text: French1444Not measuredOnly Step 5 Preview measured
LMArena Text: Spanish1445Not measuredOnly Step 5 Preview measured
LMArena Text: Russian1432Not measuredOnly Step 5 Preview measured
LMArena Text: Polish1406Not measuredOnly Step 5 Preview measured
LMArena Text: multi-turn1439Not measuredOnly Step 5 Preview measured
LMArena Text: longer queries1454Not measuredOnly Step 5 Preview measured
LMArena Text: writing and language1431Not measuredOnly Step 5 Preview measured
LMArena Text: entertainment and media1414Not measuredOnly Step 5 Preview measured
LMArena Vision: English1275Not measuredOnly Step 5 Preview measured
LMArena Vision: Chinese1318Not measuredOnly Step 5 Preview measured
LMArena Vision: reading text1279Not measuredOnly Step 5 Preview measured
LMArena Vision: humor1215Not measuredOnly Step 5 Preview measured
LMArena Vision: creative writing1250Not measuredOnly Step 5 Preview measured
Neither model has a result on the other 21 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Step 5 Preview and Hy4 flash
BenchmarkStep 5 PreviewHy4 flashLeader
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

Step 5 Preview$0.0024Fits in one request
Hy4 flash$0.0001Fits in one request

Hy4 flash costs less

Repository review

50K input + 3K output tokens

Step 5 Preview$0.058Fits in one request
Hy4 flash$0.0022Fits in one request

Hy4 flash costs less

Agent loop

200K input + 10K output tokens

Step 5 Preview$0.227Fits in one request
Hy4 flash$0.0084Fits in one request

Hy4 flash 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.

Specifications of Step 5 Preview and Hy4 flash
SpecificationStepFun Step 5 PreviewTencent Hy4 flash
MakerStepFunTencent
WeightsOpen weightsOpen weights
ReleasedOctober 8, 2026November 16, 2026
Context window1M1.05M
Input $/Mtok$1.00$0.035
Output $/Mtok$2.70$0.1405
Coding index57.344.5
Independent measurements behind it42
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Step 5 Preview better than Hy4 flash for coding?

Pick Step 5 Preview for coding: it scores 57.3 on the Vibe Coding Index against 44.5 for Hy4 flash. Pick Hy4 flash for output tokens 19.2x cheaper ($0.1405 vs $2.70 per million) and a larger context window (1.05M against 1M tokens).

Which is cheaper, Step 5 Preview or Hy4 flash?

Step 5 Preview costs $1.00 per million input tokens and $2.70 per million output tokens; Hy4 flash costs $0.035 and $0.1405. Cost per solved task is known for neither of them.

Which has the bigger context window, Step 5 Preview or Hy4 flash?

Hy4 flash, with 1.05M tokens against 1M.

How many benchmarks is the Step 5 Preview vs Hy4 flash comparison based on?

7 benchmarks have a result for both models as of November 22, 2026. Step 5 Preview's coding index rests on 4 independent benchmarks and Hy4 flash's on 2. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/stepfun-step-5-preview-vs-tencent-hy4-flash" title="Step 5 Preview vs Hy4 flash, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/stepfun-step-5-preview-vs-tencent-hy4-flash">LLMBenchmark.ai, Step 5 Preview vs Hy4 flash</a></p>