GPT-5.5 vs Hy3 preview
Updated October 21, 2026. 36 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
vs boards won of 36 shared
Decision reading
Pick GPT-5.5 for coding: it scores 67.8 on the Vibe Coding Index against 22.4 for Hy3 preview.
Pick Hy3 preview when output tokens cost 50.0x less.
Model AGPT-5.5OpenAI67.8/1002 independent benchmarks$1.65 per solved taskModel BHy3 previewTencent22.4/1001 independent benchmarkNo cost per solved task
- Shared results
- 36
- GPT-5.5 only
- 96
- Hy3 preview only
- 0
- Like-for-like categories
- 7 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-5.5
GPT-5.5 scores 67.8 against 22.4.
2 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
GPT-5.5
GPT-5.5 leads, 38.4 against 25.3.
Same board, both measured
Agentic work
AA Agentic Index
GPT-5.5
GPT-5.5 leads, 36.4 against 24.1.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only GPT-5.5 has a GPQA Diamond result, 93.2%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.5 has a MMLU Pro result, 88.1%.
Directional only
Web design
Design Arena: websites
No call
Only GPT-5.5 has a Design Arena: websites result, 1256.
Directional only
Long documents
Listed context window
GPT-5.5
GPT-5.5 takes 1.05M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Hy3 preview
GPT-5.5 $0.020, Hy3 preview $0.0005 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.5 has the number, $1.65.
Directional only
At a glance
- Coding index
- GPT-5.5 67.8 from 2 benchmarks, Hy3 preview 22.4 from 1 benchmark.
- Widest category
- Coding, 20.8 points apart on average over 7 shared benchmarks.
- Widest single gap
- LMArena WebDev: HTML: GPT-5.5 1543 against Hy3 preview 1356, 31.2 points apart on a 0 to 100 scale.
- Output price
- GPT-5.5 $30.00 against Hy3 preview $0.60 per million output tokens, 50.0x apart.
- Context window
- GPT-5.5 takes 1.05M tokens, Hy3 preview 262K.
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-5.5 | Bars | Hy3 preview | Ties |
|---|---|---|---|---|
| Coding | 7 | 0 | ||
| Intelligence | 1 | 0 | ||
| Agentic work | 2 | 0 | ||
| Reasoning | 5 | 0 | ||
| Math | 2 | 0 | ||
| Knowledge | 4 | 0 | ||
| General | 15 | 0 | ||
| All shared boards | 36 | 0 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index70
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| LiveBench coding | 85.1% | Not measured | Only GPT-5.5 measured |
| LMArena WebDev | 1513 | 1357 | GPT-5.5 |
| AA Coding Index | 74.9 | 58.8 | GPT-5.5 |
| LiveCodeBench (Vals AI) | 85.3% | Not measured | Only GPT-5.5 measured |
| SWE-bench Verified (Vals AI) | 82.6% | Not measured | Only GPT-5.5 measured |
| Vibe Code Bench (Vals AI) | 69.8% | Not measured | Only GPT-5.5 measured |
| Code migration (Vals AI) | 45.2% | Not measured | Only GPT-5.5 measured |
| ProgramBench (Vals AI) | 0.5% | Not measured | Only GPT-5.5 measured |
| Terminal-Bench 2.1 (Vals AI) | 76.4% | Not measured | Only GPT-5.5 measured |
| SkillsBench (Vals AI) | 62.5% | Not measured | Only GPT-5.5 measured |
| SciCode | 56.1% | Not measured | Only GPT-5.5 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 60.6% | 34.1% | GPT-5.5 |
| Terminal-Bench | 84.7% | Not measured | Only GPT-5.5 measured |
| FrontierCode | 43.0% | Not measured | Only GPT-5.5 measured |
| GSO-Bench | 40.2% | Not measured | Only GPT-5.5 measured |
| WeirdML | 84.9% | Not measured | Only GPT-5.5 measured |
| MirrorCode | 10.0% | Not measured | Only GPT-5.5 measured |
| ALE-Bench | 1943.0 | Not measured | Only GPT-5.5 measured |
| LMArena Text: coding | 1494 | 1421 | GPT-5.5 |
| LMArena Text: software and IT | 1493 | 1423 | GPT-5.5 |
| LMArena WebDev: React | 1506 | 1349 | GPT-5.5 |
| LMArena WebDev: HTML | 1543 | 1356 | GPT-5.5 |
| LMArena WebDev: from an image | 1532 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 15 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| AA Intelligence Index | 38.4 | 25.3 | GPT-5.5 |
| Epoch Capabilities Index | 159.7 | Not measured | Only GPT-5.5 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| AA Agentic Index | 36.4 | 24.1 | GPT-5.5 |
| Design Arena: agents building web apps | 1117 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building full-stack apps | 1089 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building mobile apps | 1143 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building Android apps | 1169 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building Godot games | 1173 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building games | 1179 | Not measured | Only GPT-5.5 measured |
| Design Arena: HTML slides | 1067 | Not measured | Only GPT-5.5 measured |
| Design Arena: PowerPoint slides | 1108 | Not measured | Only GPT-5.5 measured |
| LiveBench research | 61.6% | Not measured | Only GPT-5.5 measured |
| LiveBench browser | 48.0% | Not measured | Only GPT-5.5 measured |
| LiveBench long agentic | 44.6% | Not measured | Only GPT-5.5 measured |
| SREBench (Vals AI) | 3.8% | Not measured | Only GPT-5.5 measured |
| Finance Agent (Vals AI) | 51.8% | Not measured | Only GPT-5.5 measured |
| Tax Agent Bench (Vals AI) | 21.8% | Not measured | Only GPT-5.5 measured |
| Legal Research Bench (Vals AI) | 40.4% | Not measured | Only GPT-5.5 measured |
| Excel Modeling Benchmark (Vals AI) | 64.5% | Not measured | Only GPT-5.5 measured |
| Time Horizon Index (Vals AI) | 8.3% | Not measured | Only GPT-5.5 measured |
| τ²-Bench Telecom | 93.9% | 92.7% | GPT-5.5 |
| APEX-Agents | 55.1% | Not measured | Only GPT-5.5 measured |
| Remote Labor Index | 6.3% | Not measured | Only GPT-5.5 measured |
| OSWorld 2.0 | 13.0% | Not measured | Only GPT-5.5 measured |
| DeepResearch Bench | 54.0% | Not measured | Only GPT-5.5 measured |
| PostTrainBench | 27.2% | Not measured | Only GPT-5.5 measured |
| Vending-Bench 2 | $7,524 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 14 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| GPQA Diamond | 93.2% | Not measured | Only GPT-5.5 measured |
| LiveBench reasoning | 89.7% | Not measured | Only GPT-5.5 measured |
| MMMU Pro (Vals AI) | 88.3% | Not measured | Only GPT-5.5 measured |
| GPQA Diamond (Artificial Analysis run) | 93.5% | 86.7% | GPT-5.5 |
| Humanity's Last Exam (Artificial Analysis run) | 45.8% | 27.8% | GPT-5.5 |
| GPQA Diamond (Epoch AI run) | 90.7% | Not measured | Only GPT-5.5 measured |
| ARC-AGI | 95.0% | Not measured | Only GPT-5.5 measured |
| ARC-AGI-2 | 85.0% | Not measured | Only GPT-5.5 measured |
| SimpleBench | 69.0% | Not measured | Only GPT-5.5 measured |
| Chess puzzles | 63.0% | Not measured | Only GPT-5.5 measured |
| Mystery game puzzles | 56.0% | Not measured | Only GPT-5.5 measured |
| EBR-bench | 34.3% | Not measured | Only GPT-5.5 measured |
| LMArena Text: hard prompts | 1489 | 1413 | GPT-5.5 |
| LMArena Text: expert | 1507 | 1422 | GPT-5.5 |
| LMArena Text: hard prompts in English | 1483 | 1416 | GPT-5.5 |
| LMArena Vision: diagrams | 1325 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: homework | 1332 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 4 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| OTIS Mock AIME | 84.4% | Not measured | Only GPT-5.5 measured |
| FrontierMath, tiers 1 to 3 | 85.3% | Not measured | Only GPT-5.5 measured |
| FrontierMath, tier 4 | 72.5% | Not measured | Only GPT-5.5 measured |
| ProofBench | 50.0% | Not measured | Only GPT-5.5 measured |
| LMArena Text: math | 1487 | 1424 | GPT-5.5 |
| LMArena Text: mathematical work | 1496 | 1442 | GPT-5.5 |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| MMLU Pro | 88.1% | Not measured | Only GPT-5.5 measured |
| LegalBench (Vals AI) | 86.5% | Not measured | Only GPT-5.5 measured |
| CaseLaw (Vals AI) | 66.2% | Not measured | Only GPT-5.5 measured |
| CorpFin (Vals AI) | 68.4% | Not measured | Only GPT-5.5 measured |
| TaxEval (Vals AI) | 75.0% | Not measured | Only GPT-5.5 measured |
| MortgageTax (Vals AI) | 68.8% | Not measured | Only GPT-5.5 measured |
| MedCode (Vals AI) | 49.1% | Not measured | Only GPT-5.5 measured |
| SimpleQA Verified | 63.0% | Not measured | Only GPT-5.5 measured |
| LMArena Text: science | 1485 | 1416 | GPT-5.5 |
| LMArena Text: medicine and healthcare | 1465 | 1405 | GPT-5.5 |
| LMArena Text: legal and government | 1491 | 1430 | GPT-5.5 |
| LMArena Text: business and finance | 1478 | 1397 | GPT-5.5 |
| Neither model has a result on the other 5 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| Design Arena: websites | 1256 | Not measured | Only GPT-5.5 measured |
| Design Arena: UI components | 1260 | Not measured | Only GPT-5.5 measured |
| Design Arena: game dev | 1294 | Not measured | Only GPT-5.5 measured |
| Design Arena: data viz | 1265 | Not measured | Only GPT-5.5 measured |
| Design Arena: 3D | 1198 | Not measured | Only GPT-5.5 measured |
| Design Arena: all code categories | 1260 | Not measured | Only GPT-5.5 measured |
| Design Arena: SVG | 1234 | Not measured | Only GPT-5.5 measured |
| Design Arena: ASCII art | 1254 | Not measured | Only GPT-5.5 measured |
| LiveBench design | 70.3% | Not measured | Only GPT-5.5 measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index150
| Benchmark | GPT-5.5 | Hy3 preview | Leader |
|---|---|---|---|
| LiveBench data analysis | 81.6% | Not measured | Only GPT-5.5 measured |
| Public Benefits Bench (Vals AI) | 60.9% | Not measured | Only GPT-5.5 measured |
| SAGE (Vals AI) | 51.5% | Not measured | Only GPT-5.5 measured |
| MedScribe (Vals AI) | 86.9% | Not measured | Only GPT-5.5 measured |
| Vals Multimodal Index | 68.1% | Not measured | Only GPT-5.5 measured |
| IFBench | 75.9% | 63.1% | GPT-5.5 |
| Long context reasoning (AA-LCR) | 84.3% | 64.7% | GPT-5.5 |
| DTBench | 96.0% | Not measured | Only GPT-5.5 measured |
| LMCA | 54.3% | Not measured | Only GPT-5.5 measured |
| Surface Evolver Bench | 88.1% | Not measured | Only GPT-5.5 measured |
| Furniture assembly | 44.2% | Not measured | Only GPT-5.5 measured |
| ExploitBench | 47.4% | Not measured | Only GPT-5.5 measured |
| CL-bench Life | 22.2% | Not measured | Only GPT-5.5 measured |
| LMArena Text | 1471 | 1401 | GPT-5.5 |
| LMArena Text: instruction following | 1478 | 1383 | GPT-5.5 |
| LMArena Text: creative writing | 1455 | 1339 | GPT-5.5 |
| LMArena Vision | 1297 | Not measured | Only GPT-5.5 measured |
| LMArena Text: English | 1470 | 1409 | GPT-5.5 |
| LMArena Text: non-English | 1467 | 1390 | GPT-5.5 |
| LMArena Text: Chinese | 1534 | 1435 | GPT-5.5 |
| LMArena Text: German | 1480 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Japanese | 1499 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Korean | 1460 | Not measured | Only GPT-5.5 measured |
| LMArena Text: French | 1485 | 1425 | GPT-5.5 |
| LMArena Text: Spanish | 1468 | 1399 | GPT-5.5 |
| LMArena Text: Russian | 1473 | 1399 | GPT-5.5 |
| LMArena Text: Polish | 1465 | Not measured | Only GPT-5.5 measured |
| LMArena Text: multi-turn | 1476 | 1411 | GPT-5.5 |
| LMArena Text: longer queries | 1484 | 1402 | GPT-5.5 |
| LMArena Text: writing and language | 1466 | 1370 | GPT-5.5 |
| LMArena Text: entertainment and media | 1449 | 1352 | GPT-5.5 |
| LMArena Vision: English | 1292 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: Chinese | 1360 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: reading text | 1306 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: humor | 1293 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: recognizing things | 1231 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: creative writing | 1301 | Not measured | Only GPT-5.5 measured |
| LMArena Document | 1486 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 8 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5.5 | Hy3 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
GPT-5.5$0.020Fits in one request
Hy3 preview$0.0005Fits in one request
Hy3 preview costs less
Repository review
50K input + 3K output tokens
GPT-5.5$0.340Fits in one request
Hy3 preview$0.011Fits in one request
Hy3 preview costs less
Agent loop
200K input + 10K output tokens
GPT-5.5$1.30Fits in one request
Hy3 preview$0.042Fits in one request
Hy3 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-5.5 | Tencent Hy3 preview |
|---|---|---|
| Maker | OpenAI | Tencent |
| Weights | No public weights listed | Open weights |
| Released | April 24, 2026 | April 22, 2026 |
| Context window | 1.05M | 262K |
| Input $/Mtok | $5.00 | $0.18 |
| Output $/Mtok | $30.00 | $0.60 |
| Coding index | 67.8 | 22.4 |
| Independent measurements behind it | 2 | 1 |
| Cost per solved task | $1.65 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-5.5 better than Hy3 preview for coding?
Pick GPT-5.5 for coding: it scores 67.8 on the Vibe Coding Index against 22.4 for Hy3 preview. Pick Hy3 preview when output tokens cost 50.0x less.
Which is cheaper, GPT-5.5 or Hy3 preview?
GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens; Hy3 preview costs $0.18 and $0.60. Cost per solved task is known for GPT-5.5 only.
Which has the bigger context window, GPT-5.5 or Hy3 preview?
GPT-5.5, with 1.05M tokens against 262K.
How many benchmarks is the GPT-5.5 vs Hy3 preview comparison based on?
36 benchmarks have a result for both models as of October 21, 2026. GPT-5.5's coding index rests on 2 independent benchmarks and Hy3 preview's on 1. Where only one model has a result, the page says so instead of naming a winner.