GPT-5 vs Hy4 preview
Updated October 21, 2026. 7 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick GPT-5 for coding: it scores 72.2 on the Vibe Coding Index against 61.6 for Hy4 preview.
Pick Hy4 preview when output tokens cost 4.0x less, or when the context window is larger, 1.05M against 400K tokens.
- Shared results
- 7
- GPT-5 only
- 102
- Hy4 preview only
- 17
- Like-for-like categories
- 4 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
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, four public coding benchmarks
GPT-5
GPT-5 scores 72.2 against 61.6.
3 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
No call
Only GPT-5 has a AA Intelligence Index result, 23.0.
Directional only
Reasoning
GPQA Diamond
No call
Only GPT-5 has a GPQA Diamond result, 85.6%.
Directional only
Math
AIME
No call
Only GPT-5 has a AIME result, 93.4%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5 has a MMLU Pro result, 86.5%.
Directional only
Web design
Design Arena: websites
No call
Only GPT-5 has a Design Arena: websites result, 1185.
Directional only
General
Vals Index
No call
Only Hy4 preview has a Vals Index result, 53.2%.
Directional only
Long documents
Listed context window
Hy4 preview
Hy4 preview takes 1.05M tokens against 400K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Hy4 preview
GPT-5 $0.0063, Hy4 preview $0.0021 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5 has the number, $0.39.
Directional only
At a glance
- Coding index
- GPT-5 72.2 from 3 benchmarks, Hy4 preview 61.6 from 1 benchmark.
- Widest category
- Math, 57.0 points apart on average over 1 shared benchmark.
- Widest single gap
- Vibe Code Bench (Vals AI): GPT-5 20.1% against Hy4 preview 77.5%, 57.4 points apart on a 0 to 100 scale.
- Output price
- GPT-5 $10.00 against Hy4 preview $2.501 per million output tokens, 4.0x apart.
- Context window
- GPT-5 takes 400K tokens, Hy4 preview 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.
| Category | GPT-5 | Bars | Hy4 preview | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Math | 0 | 1 | ||
| Knowledge | 2 | 0 | ||
| General | 1 | 0 | ||
| All shared boards | 3 | 4 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index03
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| Aider polyglot | 88.0% | Not measured | Only GPT-5 measured |
| SWE-bench Verified | 74.4% | Not measured | Only GPT-5 measured |
| LMArena WebDev | 1418 | 1631 | Hy4 preview |
| AA Coding Index | 37.8 | Not measured | Only GPT-5 measured |
| LiveCodeBench (Vals AI) | 85.9% | Not measured | Only GPT-5 measured |
| SWE-bench Verified (Vals AI) | 69.0% | Not measured | Only GPT-5 measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 8.1% | Only Hy4 preview measured |
| Vibe Code Bench (Vals AI) | 20.1% | 77.5% | Hy4 preview |
| Code migration (Vals AI) | Not measured | 47.4% | Only Hy4 preview measured |
| ProgramBench (Vals AI) | Not measured | 0.0% | Only Hy4 preview measured |
| IOI (Vals AI) | Not measured | 59.3% | Only Hy4 preview measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 55.1% | Only Hy4 preview measured |
| CyberBench (Vals AI) | Not measured | 65.4% | Only Hy4 preview measured |
| LiveCodeBench (Artificial Analysis run) | 84.6% | Not measured | Only GPT-5 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 37.9% | Not measured | Only GPT-5 measured |
| Terminal-Bench | 49.6% | Not measured | Only GPT-5 measured |
| GSO-Bench | 6.9% | Not measured | Only GPT-5 measured |
| WeirdML | 60.7% | Not measured | Only GPT-5 measured |
| ALE-Bench | 807.6 | Not measured | Only GPT-5 measured |
| LMArena Text: coding | 1436 | Not measured | Only GPT-5 measured |
| LMArena Text: software and IT | 1420 | Not measured | Only GPT-5 measured |
| LMArena WebDev: React | Not measured | 1653 | Only Hy4 preview measured |
| LMArena WebDev: HTML | 1427 | 1620 | Hy4 preview |
| Neither model has a result on the other 15 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| AA Intelligence Index | 23.0 | Not measured | Only GPT-5 measured |
| Epoch Capabilities Index | 150.0 | Not measured | Only GPT-5 measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index14 boards, none shared
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| SREBench (Vals AI) | Not measured | 2.3% | Only Hy4 preview measured |
| Finance Agent (Vals AI) | Not measured | 55.1% | Only Hy4 preview measured |
| Tax Agent Bench (Vals AI) | Not measured | 32.6% | Only Hy4 preview measured |
| Legal Research Bench (Vals AI) | Not measured | 45.2% | Only Hy4 preview measured |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 8.8% | Only Hy4 preview measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 58.1% | Only Hy4 preview measured |
| BioMysteryBench (Vals AI) | Not measured | 69.3% | Only Hy4 preview measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 1.4% | Only Hy4 preview measured |
| τ²-Bench Telecom | 86.5% | Not measured | Only GPT-5 measured |
| METR time horizons | 69.6% | Not measured | Only GPT-5 measured |
| Remote Labor Index | 1.7% | Not measured | Only GPT-5 measured |
| GDPval | 34.8% | Not measured | Only GPT-5 measured |
| BALROG | 32.8% | Not measured | Only GPT-5 measured |
| DeepResearch Bench | 49.6% | Not measured | Only GPT-5 measured |
| Neither model has a result on the other 25 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index17 boards, none shared
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| GPQA Diamond | 85.6% | Not measured | Only GPT-5 measured |
| MMMU Pro (Vals AI) | 81.5% | Not measured | Only GPT-5 measured |
| GPQA Diamond (Artificial Analysis run) | 85.4% | Not measured | Only GPT-5 measured |
| Humanity's Last Exam (Artificial Analysis run) | 28.5% | Not measured | Only GPT-5 measured |
| GPQA Diamond (Epoch AI run) | 86.2% | Not measured | Only GPT-5 measured |
| ARC-AGI | 65.7% | Not measured | Only GPT-5 measured |
| ARC-AGI-2 | 9.9% | Not measured | Only GPT-5 measured |
| SimpleBench | 56.7% | Not measured | Only GPT-5 measured |
| Chess puzzles | 36.0% | Not measured | Only GPT-5 measured |
| Mystery game puzzles | 23.0% | Not measured | Only GPT-5 measured |
| EBR-bench | 12.7% | Not measured | Only GPT-5 measured |
| VPCT | 66.0% | Not measured | Only GPT-5 measured |
| LMArena Text: hard prompts | 1415 | Not measured | Only GPT-5 measured |
| LMArena Text: expert | 1419 | Not measured | Only GPT-5 measured |
| LMArena Text: hard prompts in English | 1419 | Not measured | Only GPT-5 measured |
| LMArena Vision: diagrams | 1252 | Not measured | Only GPT-5 measured |
| LMArena Vision: homework | 1278 | Not measured | Only GPT-5 measured |
| Neither model has a result on the other 4 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| AIME | 93.4% | Not measured | Only GPT-5 measured |
| MATH 500 (Vals AI) | 96.0% | Not measured | Only GPT-5 measured |
| MGSM (Vals AI) | 92.8% | Not measured | Only GPT-5 measured |
| AA Math Index | 94.3 | Not measured | Only GPT-5 measured |
| AIME 2025 (Artificial Analysis run) | 94.3% | Not measured | Only GPT-5 measured |
| OTIS Mock AIME | 91.4% | Not measured | Only GPT-5 measured |
| FrontierMath, tiers 1 to 3 | 55.4% | Not measured | Only GPT-5 measured |
| FrontierMath, tier 4 | 22.0% | Not measured | Only GPT-5 measured |
| ProofBench | 18.0% | 75.0% | Hy4 preview |
| MATH level 5 | 98.1% | Not measured | Only GPT-5 measured |
| LMArena Text: math | 1407 | Not measured | Only GPT-5 measured |
| LMArena Text: mathematical work | 1405 | Not measured | Only GPT-5 measured |
| Neither model has a result on the other 1 board here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| MMLU Pro | 86.5% | Not measured | Only GPT-5 measured |
| LegalBench (Vals AI) | 86.0% | 83.8% | GPT-5 |
| CaseLaw (Vals AI) | 66.5% | Not measured | Only GPT-5 measured |
| CorpFin (Vals AI) | 61.1% | Not measured | Only GPT-5 measured |
| TaxEval (Vals AI) | 73.4% | Not measured | Only GPT-5 measured |
| MortgageTax (Vals AI) | 65.5% | Not measured | Only GPT-5 measured |
| MedQA (Vals AI) | 96.3% | Not measured | Only GPT-5 measured |
| MedCode (Vals AI) | 49.6% | 43.3% | GPT-5 |
| MMLU Pro (Artificial Analysis run) | 87.1% | Not measured | Only GPT-5 measured |
| SimpleQA Verified | 50.1% | Not measured | Only GPT-5 measured |
| GeoBench | 81.0% | Not measured | Only GPT-5 measured |
| LMArena Text: science | 1416 | Not measured | Only GPT-5 measured |
| LMArena Text: medicine and healthcare | 1413 | Not measured | Only GPT-5 measured |
| LMArena Text: legal and government | 1431 | Not measured | Only GPT-5 measured |
| LMArena Text: business and finance | 1400 | Not measured | Only GPT-5 measured |
| Neither model has a result on the other 2 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| Design Arena: websites | 1185 | Not measured | Only GPT-5 measured |
| Design Arena: UI components | 1179 | Not measured | Only GPT-5 measured |
| Design Arena: game dev | 1185 | Not measured | Only GPT-5 measured |
| Design Arena: data viz | 1221 | Not measured | Only GPT-5 measured |
| Design Arena: 3D | 1056 | Not measured | Only GPT-5 measured |
| Design Arena: all code categories | 1168 | Not measured | Only GPT-5 measured |
| Design Arena: SVG | 1188 | Not measured | Only GPT-5 measured |
| Design Arena: ASCII art | 1140 | Not measured | Only GPT-5 measured |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-5 | Hy4 preview | Leader |
|---|---|---|---|
| Vals Index | Not measured | 53.2% | Only Hy4 preview measured |
| Public Benefits Bench (Vals AI) | Not measured | 68.6% | Only Hy4 preview measured |
| SAGE (Vals AI) | 43.7% | Not measured | Only GPT-5 measured |
| MedScribe (Vals AI) | 83.7% | 83.6% | GPT-5 |
| IFBench | 73.1% | Not measured | Only GPT-5 measured |
| Long context reasoning (AA-LCR) | 78.2% | Not measured | Only GPT-5 measured |
| DTBench | 90.7% | Not measured | Only GPT-5 measured |
| LMCA | 40.0% | Not measured | Only GPT-5 measured |
| Fiction.LiveBench | 97.2% | Not measured | Only GPT-5 measured |
| Creative writing (Lech Mazur) | 86.0% | Not measured | Only GPT-5 measured |
| LMArena Text | 1406 | Not measured | Only GPT-5 measured |
| LMArena Text: instruction following | 1388 | Not measured | Only GPT-5 measured |
| LMArena Text: creative writing | 1365 | Not measured | Only GPT-5 measured |
| LMArena Vision | 1232 | Not measured | Only GPT-5 measured |
| LMArena Text: English | 1409 | Not measured | Only GPT-5 measured |
| LMArena Text: non-English | 1397 | Not measured | Only GPT-5 measured |
| LMArena Text: Chinese | 1422 | Not measured | Only GPT-5 measured |
| LMArena Text: German | 1416 | Not measured | Only GPT-5 measured |
| LMArena Text: Japanese | 1409 | Not measured | Only GPT-5 measured |
| LMArena Text: Korean | 1360 | Not measured | Only GPT-5 measured |
| LMArena Text: French | 1409 | Not measured | Only GPT-5 measured |
| LMArena Text: Spanish | 1399 | Not measured | Only GPT-5 measured |
| LMArena Text: Russian | 1407 | Not measured | Only GPT-5 measured |
| LMArena Text: Polish | 1431 | Not measured | Only GPT-5 measured |
| LMArena Text: multi-turn | 1426 | Not measured | Only GPT-5 measured |
| LMArena Text: longer queries | 1399 | Not measured | Only GPT-5 measured |
| LMArena Text: writing and language | 1384 | Not measured | Only GPT-5 measured |
| LMArena Text: entertainment and media | 1379 | Not measured | Only GPT-5 measured |
| LMArena Vision: English | 1231 | Not measured | Only GPT-5 measured |
| LMArena Vision: Chinese | 1267 | Not measured | Only GPT-5 measured |
| LMArena Vision: reading text | 1246 | Not measured | Only GPT-5 measured |
| LMArena Vision: humor | 1219 | Not measured | Only GPT-5 measured |
| LMArena Vision: captioning | 1212 | Not measured | Only GPT-5 measured |
| LMArena Vision: recognizing things | 1263 | Not measured | Only GPT-5 measured |
| LMArena Vision: creative writing | 1210 | Not measured | Only GPT-5 measured |
| Neither model has a result on the other 11 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5 | Hy4 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
Hy4 preview costs less
Repository review
50K input + 3K output tokens
Hy4 preview costs less
Agent loop
200K input + 10K output tokens
Hy4 preview 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 | OpenAI GPT-5 | Tencent Hy4 preview |
|---|---|---|
| Maker | OpenAI | Tencent |
| Weights | No public weights listed | Open weights |
| Released | August 7, 2025 | August 28, 2026 |
| Context window | 400K | 1.05M |
| Input $/Mtok | $1.25 | $0.834 |
| Output $/Mtok | $10.00 | $2.501 |
| Coding index | 72.2 | 61.6 |
| Independent measurements behind it | 3 | 1 |
| Cost per solved task | $0.39 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-5 better than Hy4 preview for coding?
Pick GPT-5 for coding: it scores 72.2 on the Vibe Coding Index against 61.6 for Hy4 preview. Pick Hy4 preview when output tokens cost 4.0x less, or when the context window is larger, 1.05M against 400K tokens.
Which is cheaper, GPT-5 or Hy4 preview?
GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; Hy4 preview costs $0.834 and $2.501. Cost per solved task is known for GPT-5 only.
Which has the bigger context window, GPT-5 or Hy4 preview?
Hy4 preview, with 1.05M tokens against 400K.
How many benchmarks is the GPT-5 vs Hy4 preview comparison based on?
7 benchmarks have a result for both models as of October 21, 2026. GPT-5's coding index rests on 3 independent benchmarks and Hy4 preview's on 1. Where only one model has a result, the page says so instead of naming a winner.