Enter opens the highlighted result. Esc closes.

GPT-5.2 vs Hy4 flash

GPT-5.2 leads Hy4 flash on coding, 56.0 to 44.5 on the Vibe Coding Index, and Hy4 flash costs 99.6x less per output token ($0.1405 vs $14.00 per million). Updated November 22, 2026; 6 benchmarks have a result for both.

GPT-5.2OpenAI
Coding
56.0
Overall
64.4
Output $/M
$14.00
0
Hy4 flashTencent
Coding
44.5
Overall
No data
Output $/M
$0.1405
6
vs boards won of 6 shared

GPT-5.2 or Hy4 flash: which to pick

Pick GPT-5.2 for coding: it scores 56.0 on the Vibe Coding Index against 44.5 for Hy4 flash.

Pick Hy4 flash for output tokens 99.6x cheaper ($0.1405 vs $14.00 per million) and a larger context window (1.05M against 400K tokens).

Shared results
6
GPT-5.2 only
108
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
GPT-5.2 56.0 from 3 benchmarks, Hy4 flash 44.5 from 2 benchmarks.
Widest category
Coding, 21.3 points apart on average over 1 shared benchmark.
Widest single gap
Vibe Code Bench (Vals AI): GPT-5.2 53.5% against Hy4 flash 74.8%, 21.3 points apart on a 0 to 100 scale.
Output price
GPT-5.2 $14.00 against Hy4 flash $0.1405 per million output tokens, 99.6x apart.
Context window
GPT-5.2 takes 400K 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
CategoryGPT-5.2BarsHy4 flashTies
Coding01
Web design05
All shared boards06
CodingLike-for-likeSix of these boards feed the Vibe Coding Index01
Coding benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
SWE-bench Verified72.8%Not measuredOnly GPT-5.2 measured
SWE-bench Multilingual66.7%Not measuredOnly GPT-5.2 measured
LMArena WebDev1419Not measuredOnly GPT-5.2 measured
LiveCodeBench (Vals AI)85.4%Not measuredOnly GPT-5.2 measured
SWE-bench Verified (Vals AI)75.8%Not measuredOnly GPT-5.2 measured
Terminal-Bench 4 (Vals AI)Not measured15.2%Only Hy4 flash measured
Vibe Code Bench (Vals AI)53.5%74.8%Hy4 flash
LiveCodeBench (Artificial Analysis run)88.9%Not measuredOnly GPT-5.2 measured
Terminal-Bench Hard (Artificial Analysis run)47.0%Not measuredOnly GPT-5.2 measured
Terminal-Bench64.9%Not measuredOnly GPT-5.2 measured
GSO-Bench27.5%Not measuredOnly GPT-5.2 measured
WeirdML72.2%Not measuredOnly GPT-5.2 measured
MirrorCode32.2%Not measuredOnly GPT-5.2 measured
ALE-Bench1293.5Not measuredOnly GPT-5.2 measured
LMArena Text: coding1442Not measuredOnly GPT-5.2 measured
LMArena Text: software and IT1434Not measuredOnly GPT-5.2 measured
LMArena WebDev: HTML1429Not measuredOnly GPT-5.2 measured
GameBench 2Not measured50.5Only Hy4 flash measured
Terminal-Bench 2.064.9%Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 33 boards here
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
Intelligence benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
Epoch Capabilities Index154.5Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 1 board here
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index16 boards, none shared
Agentic work benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
Design Arena: agents building web apps1054Not measuredOnly GPT-5.2 measured
Design Arena: agents building full-stack apps1020Not measuredOnly GPT-5.2 measured
Design Arena: agents building mobile apps1086Not measuredOnly GPT-5.2 measured
Design Arena: agents building Android apps1043Not measuredOnly GPT-5.2 measured
Design Arena: agents building Godot games1142Not measuredOnly GPT-5.2 measured
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 Telecom84.8%Not measuredOnly GPT-5.2 measured
METR time horizons75.3%Not measuredOnly GPT-5.2 measured
Remote Labor Index2.5%Not measuredOnly GPT-5.2 measured
GDPval49.7%Not measuredOnly GPT-5.2 measured
DeepResearch Bench41.1%Not measuredOnly GPT-5.2 measured
Vending-Bench 2$3,591Not measuredOnly GPT-5.2 measured
Berkeley Function Calling Leaderboard (BFCL)55.9%Not measuredOnly GPT-5.2 measured
τ³-Banking32.2%Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 28 boards here
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index18 boards, none shared
Reasoning benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
GPQA Diamond91.7%Not measuredOnly GPT-5.2 measured
MMMU Pro (Vals AI)86.7%Not measuredOnly GPT-5.2 measured
GPQA Diamond (Artificial Analysis run)90.3%Not measuredOnly GPT-5.2 measured
Humanity's Last Exam (Artificial Analysis run)37.7%Not measuredOnly GPT-5.2 measured
GPQA Diamond (Epoch AI run)91.4%Not measuredOnly GPT-5.2 measured
ARC-AGI86.2%Not measuredOnly GPT-5.2 measured
ARC-AGI-252.9%Not measuredOnly GPT-5.2 measured
SimpleBench45.8%Not measuredOnly GPT-5.2 measured
Chess puzzles48.0%Not measuredOnly GPT-5.2 measured
Mystery game puzzles23.0%Not measuredOnly GPT-5.2 measured
EBR-bench23.0%Not measuredOnly GPT-5.2 measured
VPCT84.0%Not measuredOnly GPT-5.2 measured
LMArena Text: hard prompts1428Not measuredOnly GPT-5.2 measured
LMArena Text: expert1444Not measuredOnly GPT-5.2 measured
LMArena Text: hard prompts in English1431Not measuredOnly GPT-5.2 measured
LMArena Vision: diagrams1265Not measuredOnly GPT-5.2 measured
LMArena Vision: homework1276Not measuredOnly GPT-5.2 measured
Kagi LLM Benchmark73.3%Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 8 boards here
MathDirectional onlyPublic boards, shown as published, kept out of the index10 boards, none shared
Math benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
AIME96.9%Not measuredOnly GPT-5.2 measured
MGSM (Vals AI)94.0%Not measuredOnly GPT-5.2 measured
AIME 2025 (Artificial Analysis run)99.0%Not measuredOnly GPT-5.2 measured
OTIS Mock AIME96.1%Not measuredOnly GPT-5.2 measured
FrontierMath, tiers 1 to 367.4%Not measuredOnly GPT-5.2 measured
FrontierMath, tier 431.7%Not measuredOnly GPT-5.2 measured
ProofBench15.0%Not measuredOnly GPT-5.2 measured
LMArena Text: math1440Not measuredOnly GPT-5.2 measured
LMArena Text: mathematical work1442Not measuredOnly GPT-5.2 measured
MathArena30.7%Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 4 boards here
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index14 boards, none shared
Knowledge benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
MMLU Pro86.2%Not measuredOnly GPT-5.2 measured
LegalBench (Vals AI)82.8%Not measuredOnly GPT-5.2 measured
CaseLaw (Vals AI)66.0%Not measuredOnly GPT-5.2 measured
CorpFin (Vals AI)65.9%Not measuredOnly GPT-5.2 measured
TaxEval (Vals AI)75.8%Not measuredOnly GPT-5.2 measured
MortgageTax (Vals AI)67.1%Not measuredOnly GPT-5.2 measured
MedQA (Vals AI)94.1%Not measuredOnly GPT-5.2 measured
MedCode (Vals AI)49.8%Not measuredOnly GPT-5.2 measured
MMLU Pro (Artificial Analysis run)87.4%Not measuredOnly GPT-5.2 measured
SimpleQA Verified37.1%Not measuredOnly GPT-5.2 measured
LMArena Text: science1425Not measuredOnly GPT-5.2 measured
LMArena Text: medicine and healthcare1425Not measuredOnly GPT-5.2 measured
LMArena Text: legal and government1428Not measuredOnly GPT-5.2 measured
LMArena Text: business and finance1414Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 7 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index05
Web design benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
Design Arena: websites11821227Hy4 flash
Design Arena: UI components11741262Hy4 flash
Design Arena: game dev11671202Hy4 flash
Design Arena: data viz1202Not measuredOnly GPT-5.2 measured
Design Arena: 3D10461209Hy4 flash
Design Arena: all code categories11511225Hy4 flash
Design Arena: SVG1123Not measuredOnly GPT-5.2 measured
Design Arena: ASCII art1138Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 1 board here
GeneralDirectional onlyPublic boards, shown as published, kept out of the index34 boards, none shared
General benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 flashLeader
Vals IndexNot measured50.8%Only Hy4 flash measured
MedScribe (Vals AI)84.4%Not measuredOnly GPT-5.2 measured
IFBench75.4%Not measuredOnly GPT-5.2 measured
Long context reasoning (AA-LCR)82.7%Not measuredOnly GPT-5.2 measured
DTBench90.9%Not measuredOnly GPT-5.2 measured
LMCA43.9%Not measuredOnly GPT-5.2 measured
CL-bench18.2%Not measuredOnly GPT-5.2 measured
Furniture assembly38.3%Not measuredOnly GPT-5.2 measured
LMArena Text1416Not measuredOnly GPT-5.2 measured
LMArena Text: instruction following1409Not measuredOnly GPT-5.2 measured
LMArena Text: creative writing1375Not measuredOnly GPT-5.2 measured
LMArena Vision1246Not measuredOnly GPT-5.2 measured
LMArena Text: English1422Not measuredOnly GPT-5.2 measured
LMArena Text: non-English1405Not measuredOnly GPT-5.2 measured
LMArena Text: Chinese1448Not measuredOnly GPT-5.2 measured
LMArena Text: German1442Not measuredOnly GPT-5.2 measured
LMArena Text: Japanese1419Not measuredOnly GPT-5.2 measured
LMArena Text: Korean1375Not measuredOnly GPT-5.2 measured
LMArena Text: French1454Not measuredOnly GPT-5.2 measured
LMArena Text: Spanish1412Not measuredOnly GPT-5.2 measured
LMArena Text: Russian1415Not measuredOnly GPT-5.2 measured
LMArena Text: Polish1425Not measuredOnly GPT-5.2 measured
LMArena Text: multi-turn1419Not measuredOnly GPT-5.2 measured
LMArena Text: longer queries1412Not measuredOnly GPT-5.2 measured
LMArena Text: writing and language1392Not measuredOnly GPT-5.2 measured
LMArena Text: entertainment and media1378Not measuredOnly GPT-5.2 measured
LMArena Vision: English1244Not measuredOnly GPT-5.2 measured
LMArena Vision: Chinese1305Not measuredOnly GPT-5.2 measured
LMArena Vision: reading text1257Not measuredOnly GPT-5.2 measured
LMArena Vision: humor1227Not measuredOnly GPT-5.2 measured
LMArena Vision: captioning1265Not measuredOnly GPT-5.2 measured
LMArena Vision: recognizing things1232Not measuredOnly GPT-5.2 measured
LMArena Vision: creative writing1226Not measuredOnly GPT-5.2 measured
LMArena Document1404Not measuredOnly GPT-5.2 measured
Neither model has a result on the other 15 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-5.2 and Hy4 flash
BenchmarkGPT-5.2Hy4 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

GPT-5.2$0.0088Fits in one request
Hy4 flash$0.0001Fits in one request

Hy4 flash costs less

Repository review

50K input + 3K output tokens

GPT-5.2$0.130Fits in one request
Hy4 flash$0.0022Fits in one request

Hy4 flash costs less

Agent loop

200K input + 10K output tokens

GPT-5.2$0.490Fits 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 GPT-5.2 and Hy4 flash
SpecificationOpenAI GPT-5.2Tencent Hy4 flash
MakerOpenAITencent
WeightsNo public weights listedOpen weights
ReleasedDecember 10, 2025November 16, 2026
Context window400K1.05M
Input $/Mtok$1.75$0.035
Output $/Mtok$14.00$0.1405
Coding index56.044.5
Independent measurements behind it32
Cost per solved task$1.91Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-5.2 better than Hy4 flash for coding?

Pick GPT-5.2 for coding: it scores 56.0 on the Vibe Coding Index against 44.5 for Hy4 flash. Pick Hy4 flash for output tokens 99.6x cheaper ($0.1405 vs $14.00 per million) and a larger context window (1.05M against 400K tokens).

Which is cheaper, GPT-5.2 or Hy4 flash?

GPT-5.2 costs $1.75 per million input tokens and $14.00 per million output tokens; Hy4 flash costs $0.035 and $0.1405. Cost per solved task is known for GPT-5.2 only.

Which has the bigger context window, GPT-5.2 or Hy4 flash?

Hy4 flash, with 1.05M tokens against 400K.

How many benchmarks is the GPT-5.2 vs Hy4 flash comparison based on?

6 benchmarks have a result for both models as of November 22, 2026. GPT-5.2'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/gpt-5-2-vs-tencent-hy4-flash" title="GPT-5.2 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/gpt-5-2-vs-tencent-hy4-flash">LLMBenchmark.ai, GPT-5.2 vs Hy4 flash</a></p>