Enter opens the highlighted result. Esc closes.

DeepSeek V3.2 vs Hy4 flash

Hy4 flash leads DeepSeek V3.2 on coding, 44.5 to 30.0 on the Vibe Coding Index, and Hy4 flash costs 4.3x less per output token ($0.1405 vs $0.61 per million). Updated November 22, 2026; 6 benchmarks have a result for both.

DeepSeek V3.2DeepSeek
Coding
30.0
Overall
40.0
Output $/M
$0.61
0
Hy4 flashTencent
Coding
44.5
Overall
No data
Output $/M
$0.1405
6
vs boards won of 6 shared

DeepSeek V3.2 or Hy4 flash: which to pick

Pick Hy4 flash for coding: it scores 44.5 on the Vibe Coding Index against 30.0 for DeepSeek V3.2, though on 2 benchmarks against 3.

DeepSeek V3.2 is neither cheaper nor larger in context, and leads no benchmark both models were measured on.

Shared results
6
DeepSeek V3.2 only
69
Hy4 flash only
6
Like-for-like categories
2 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
DeepSeek V3.2 30.0 from 3 benchmarks, Hy4 flash 44.5 from 2 benchmarks.
Widest category
Coding, 69.7 points apart on average over 1 shared benchmark.
Widest single gap
Vibe Code Bench (Vals AI): DeepSeek V3.2 5.1% against Hy4 flash 74.8%, 69.7 points apart on a 0 to 100 scale.
Output price
DeepSeek V3.2 $0.61 against Hy4 flash $0.1405 per million output tokens, 4.3x apart.
Context window
DeepSeek V3.2 takes 164K 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
CategoryDeepSeek V3.2BarsHy4 flashTies
Coding01
Web design05
All shared boards06
CodingLike-for-likeSix of these boards feed the Vibe Coding Index01
Coding benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
SWE-bench Verified70.0%Not measuredOnly DeepSeek V3.2 measured
SWE-bench Multilingual59.0%Not measuredOnly DeepSeek V3.2 measured
LMArena WebDev1362Not measuredOnly DeepSeek V3.2 measured
LiveCodeBench (Vals AI)80.7%Not measuredOnly DeepSeek V3.2 measured
SWE-bench Verified (Vals AI)67.6%Not measuredOnly DeepSeek V3.2 measured
Terminal-Bench 4 (Vals AI)Not measured15.2%Only Hy4 flash measured
Vibe Code Bench (Vals AI)5.1%74.8%Hy4 flash
LiveCodeBench (Artificial Analysis run)86.2%Not measuredOnly DeepSeek V3.2 measured
Terminal-Bench Hard (Artificial Analysis run)35.6%Not measuredOnly DeepSeek V3.2 measured
Terminal-Bench39.6%Not measuredOnly DeepSeek V3.2 measured
LMArena Text: coding1454Not measuredOnly DeepSeek V3.2 measured
LMArena Text: software and IT1450Not measuredOnly DeepSeek V3.2 measured
LMArena WebDev: React1349Not measuredOnly DeepSeek V3.2 measured
LMArena WebDev: HTML1370Not measuredOnly DeepSeek V3.2 measured
GameBench 2Not measured50.5Only Hy4 flash measured
Terminal-Bench 2.039.6%Not measuredOnly DeepSeek V3.2 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
Intelligence benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
Epoch Capabilities Index146.3Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 1 board here
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
Agentic work benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
Tax Agent Bench (Vals AI)Not measured23.3%Only Hy4 flash measured
Legal Research Bench (Vals AI)Not measured38.5%Only Hy4 flash measured
Harvey's Legal Agent Benchmark (Vals AI)Not measured5.4%Only Hy4 flash measured
τ²-Bench Telecom90.6%Not measuredOnly DeepSeek V3.2 measured
APEX-Agents21.3%Not measuredOnly DeepSeek V3.2 measured
Vending-Bench 2$1,034Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 38 boards here
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
Reasoning benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
GPQA Diamond80.3%Not measuredOnly DeepSeek V3.2 measured
GPQA Diamond (Artificial Analysis run)84.0%Not measuredOnly DeepSeek V3.2 measured
Humanity's Last Exam (Artificial Analysis run)24.6%Not measuredOnly DeepSeek V3.2 measured
ARC-AGI57.0%Not measuredOnly DeepSeek V3.2 measured
ARC-AGI-24.0%Not measuredOnly DeepSeek V3.2 measured
LMArena Text: hard prompts1434Not measuredOnly DeepSeek V3.2 measured
LMArena Text: expert1435Not measuredOnly DeepSeek V3.2 measured
LMArena Text: hard prompts in English1446Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 18 boards here
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
Math benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
AIME84.6%Not measuredOnly DeepSeek V3.2 measured
MGSM (Vals AI)86.0%Not measuredOnly DeepSeek V3.2 measured
AIME 2025 (Artificial Analysis run)92.0%Not measuredOnly DeepSeek V3.2 measured
LMArena Text: math1435Not measuredOnly DeepSeek V3.2 measured
LMArena Text: mathematical work1438Not measuredOnly DeepSeek V3.2 measured
MathArena14.4%Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 8 boards here
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index11 boards, none shared
Knowledge benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
MMLU Pro84.9%Not measuredOnly DeepSeek V3.2 measured
LegalBench (Vals AI)76.1%Not measuredOnly DeepSeek V3.2 measured
CaseLaw (Vals AI)55.4%Not measuredOnly DeepSeek V3.2 measured
CorpFin (Vals AI)51.0%Not measuredOnly DeepSeek V3.2 measured
TaxEval (Vals AI)68.2%Not measuredOnly DeepSeek V3.2 measured
MedQA (Vals AI)93.9%Not measuredOnly DeepSeek V3.2 measured
MMLU Pro (Artificial Analysis run)86.2%Not measuredOnly DeepSeek V3.2 measured
LMArena Text: science1446Not measuredOnly DeepSeek V3.2 measured
LMArena Text: medicine and healthcare1440Not measuredOnly DeepSeek V3.2 measured
LMArena Text: legal and government1425Not measuredOnly DeepSeek V3.2 measured
LMArena Text: business and finance1410Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 10 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index05
Web design benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
Design Arena: websites11621227Hy4 flash
Design Arena: UI components11321262Hy4 flash
Design Arena: game dev11031202Hy4 flash
Design Arena: data viz1153Not measuredOnly DeepSeek V3.2 measured
Design Arena: 3D10991209Hy4 flash
Design Arena: all code categories11461225Hy4 flash
Design Arena: SVG1016Not measuredOnly DeepSeek V3.2 measured
Design Arena: ASCII art1068Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 1 board here
GeneralDirectional onlyPublic boards, shown as published, kept out of the index25 boards, none shared
General benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 flashLeader
Vals IndexNot measured50.8%Only Hy4 flash measured
IFBench60.7%Not measuredOnly DeepSeek V3.2 measured
Long context reasoning (AA-LCR)73.3%Not measuredOnly DeepSeek V3.2 measured
GDPval-AA698Not measuredOnly DeepSeek V3.2 measured
DTBench85.6%Not measuredOnly DeepSeek V3.2 measured
LMCA28.8%Not measuredOnly DeepSeek V3.2 measured
CL-bench12.4%Not measuredOnly DeepSeek V3.2 measured
CL-bench Life7.4%Not measuredOnly DeepSeek V3.2 measured
LMArena Text1425Not measuredOnly DeepSeek V3.2 measured
LMArena Text: instruction following1413Not measuredOnly DeepSeek V3.2 measured
LMArena Text: creative writing1400Not measuredOnly DeepSeek V3.2 measured
LMArena Text: English1439Not measuredOnly DeepSeek V3.2 measured
LMArena Text: non-English1409Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Chinese1455Not measuredOnly DeepSeek V3.2 measured
LMArena Text: German1412Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Japanese1379Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Korean1370Not measuredOnly DeepSeek V3.2 measured
LMArena Text: French1430Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Spanish1425Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Russian1423Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Polish1419Not measuredOnly DeepSeek V3.2 measured
LMArena Text: multi-turn1427Not measuredOnly DeepSeek V3.2 measured
LMArena Text: longer queries1428Not measuredOnly DeepSeek V3.2 measured
LMArena Text: writing and language1407Not measuredOnly DeepSeek V3.2 measured
LMArena Text: entertainment and media1395Not measuredOnly DeepSeek V3.2 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
Image generation benchmarks for DeepSeek V3.2 and Hy4 flash
BenchmarkDeepSeek V3.2Hy4 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

DeepSeek V3.2$0.0006Fits in one request
Hy4 flash$0.0001Fits in one request

Hy4 flash costs less

Repository review

50K input + 3K output tokens

DeepSeek V3.2$0.015Fits in one request
Hy4 flash$0.0022Fits in one request

Hy4 flash costs less

Agent loop

200K input + 10K output tokens

DeepSeek V3.2$0.060Exceeds the context window
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 DeepSeek V3.2 and Hy4 flash
SpecificationDeepSeek V3.2Tencent Hy4 flash
MakerDeepSeekTencent
WeightsOpen weightsOpen weights
ReleasedDecember 1, 2025November 16, 2026
Context window164K1.05M
Input $/Mtok$0.27$0.035
Output $/Mtok$0.61$0.1405
Coding index30.044.5
Independent measurements behind it32
Cost per solved task$0.99Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is DeepSeek V3.2 better than Hy4 flash for coding?

Pick Hy4 flash for coding: it scores 44.5 on the Vibe Coding Index against 30.0 for DeepSeek V3.2, though on 2 benchmarks against 3. DeepSeek V3.2 is neither cheaper nor larger in context, and leads no benchmark both models were measured on.

Which is cheaper, DeepSeek V3.2 or Hy4 flash?

DeepSeek V3.2 costs $0.27 per million input tokens and $0.61 per million output tokens; Hy4 flash costs $0.035 and $0.1405. Cost per solved task is known for DeepSeek V3.2 only.

Which has the bigger context window, DeepSeek V3.2 or Hy4 flash?

Hy4 flash, with 1.05M tokens against 164K.

How many benchmarks is the DeepSeek V3.2 vs Hy4 flash comparison based on?

6 benchmarks have a result for both models as of November 22, 2026. DeepSeek V3.2's coding index rests on 3 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/deepseek-v3-2-vs-tencent-hy4-flash" title="DeepSeek V3.2 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/deepseek-v3-2-vs-tencent-hy4-flash">LLMBenchmark.ai, DeepSeek V3.2 vs Hy4 flash</a></p>