GPT-5.5 vs Hy4 flash
GPT-5.5 leads Hy4 flash on coding, 67.9 to 44.5 on the Vibe Coding Index, and Hy4 flash costs 213.5x less per output token ($0.1405 vs $30.00 per million). Updated November 22, 2026; 9 benchmarks have a result for both.
GPT-5.5 or Hy4 flash: which to pick
Pick GPT-5.5 for coding: it scores 67.9 on the Vibe Coding Index against 44.5 for Hy4 flash.
Pick Hy4 flash for output tokens 213.5x cheaper ($0.1405 vs $30.00 per million) and work measured by Vibe Code Bench (Vals AI) or Tax Agent Bench (Vals AI), where it leads.
- Shared results
- 9
- GPT-5.5 only
- 128
- Hy4 flash only
- 3
- 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.
Coding
Vibe Coding Index, six public coding benchmarks
GPT-5.5
GPT-5.5 scores 67.9 against 44.5.
4 and 2 benchmarks behind the scores
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
GPT-5.5
GPT-5.5 leads, 1243 against 1227.
Same board, both measured
General
Vals Index
No call
Only Hy4 flash has a Vals Index result, 50.8%.
Directional only
Long documents
Listed context window
GPT-5.5
GPT-5.5 takes 1.05M tokens against 1.05M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Hy4 flash
GPT-5.5 $0.020, Hy4 flash $0.0001 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 67.9 from 4 benchmarks, Hy4 flash 44.5 from 2 benchmarks.
- Widest category
- Coding, 7.3 points apart on average over 2 shared benchmarks.
- Widest single gap
- GameBench 2: GPT-5.5 60.2 against Hy4 flash 50.5, 9.6 points apart on a 0 to 100 scale.
- Output price
- GPT-5.5 $30.00 against Hy4 flash $0.1405 per million output tokens, 213.5x apart.
- Context window
- GPT-5.5 takes 1.05M 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.
| Category | GPT-5.5 | Bars | Hy4 flash | Ties |
|---|---|---|---|---|
| Coding | 1 | 1 | ||
| Agentic work | 1 | 1 | ||
| Web design | 3 | 2 | ||
| All shared boards | 5 | 4 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index11
| Benchmark | GPT-5.5 | Hy4 flash | Leader |
|---|---|---|---|
| LMArena WebDev | 1512 | Not measured | Only GPT-5.5 measured |
| 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 |
| Terminal-Bench 4 (Vals AI) | Not measured | 15.2% | Only Hy4 flash measured |
| Vibe Code Bench (Vals AI) | 69.8% | 74.8% | Hy4 flash |
| 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% | Not measured | Only GPT-5.5 measured |
| Terminal-Bench 4.0 (Artificial Analysis run) | 14.6% | Not measured | Only GPT-5.5 measured |
| 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 | 23.3% | Not measured | Only GPT-5.5 measured |
| ALE-Bench | 1943.0 | Not measured | Only GPT-5.5 measured |
| LMArena Text: coding | 1493 | Not measured | Only GPT-5.5 measured |
| LMArena Text: software and IT | 1492 | Not measured | Only GPT-5.5 measured |
| LMArena WebDev: React | 1505 | Not measured | Only GPT-5.5 measured |
| LMArena WebDev: HTML | 1543 | Not measured | Only GPT-5.5 measured |
| LMArena WebDev: from an image | 1531 | Not measured | Only GPT-5.5 measured |
| SWE Atlas: Codebase QnA | 45.4% | Not measured | Only GPT-5.5 measured |
| SWE Atlas: Refactoring | 44.8% | Not measured | Only GPT-5.5 measured |
| SWE Atlas: Test Writing | 42.6% | Not measured | Only GPT-5.5 measured |
| GameBench 2 | 60.2 | 50.5 | GPT-5.5 |
| Terminal-Bench 2.0 | 84.7% | Not measured | Only GPT-5.5 measured |
| Terminal-Bench 2.1 | 83.2% | Not measured | Only GPT-5.5 measured |
| ProgramBench | 0.5% | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 22 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | GPT-5.5 | Hy4 flash | Leader |
|---|---|---|---|
| Epoch Capabilities Index | 160.2 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 1 board here | |||
Agentic workLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | GPT-5.5 | Hy4 flash | Leader |
|---|---|---|---|
| Design Arena: agents building web apps | 1085 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building full-stack apps | 1067 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building mobile apps | 1134 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building Android apps | 1152 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building Godot games | 1143 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building games | 1125 | Not measured | Only GPT-5.5 measured |
| Design Arena: HTML slides | 969 | Not measured | Only GPT-5.5 measured |
| Design Arena: PowerPoint slides | 1078 | 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% | 23.3% | Hy4 flash |
| Legal Research Bench (Vals AI) | 40.4% | 38.5% | GPT-5.5 |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 5.4% | Only Hy4 flash 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% | Not measured | Only GPT-5.5 measured |
| 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 | 23.9% | Not measured | Only GPT-5.5 measured |
| Vending-Bench 2 | $7,524 | Not measured | Only GPT-5.5 measured |
| LMArena Search | 1246 | Not measured | Only GPT-5.5 measured |
| MCP Atlas | 75.3% | Not measured | Only GPT-5.5 measured |
| Toolathlon-Verified | 73.5% | Not measured | Only GPT-5.5 measured |
| τ³-Banking | 44.6% | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 18 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index17 boards, none shared
| Benchmark | GPT-5.5 | Hy4 flash | Leader |
|---|---|---|---|
| GPQA Diamond | 93.2% | 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% | Not measured | Only GPT-5.5 measured |
| Humanity's Last Exam (Artificial Analysis run) | 45.8% | Not measured | Only GPT-5.5 measured |
| 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 | 1488 | Not measured | Only GPT-5.5 measured |
| LMArena Text: expert | 1507 | Not measured | Only GPT-5.5 measured |
| LMArena Text: hard prompts in English | 1483 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: diagrams | 1323 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: homework | 1332 | Not measured | Only GPT-5.5 measured |
| Kagi LLM Benchmark | 88.8% | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 9 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | GPT-5.5 | Hy4 flash | 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 | 1486 | Not measured | Only GPT-5.5 measured |
| LMArena Text: mathematical work | 1494 | Not measured | Only GPT-5.5 measured |
| MathArena | 54.7% | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index12 boards, none shared
| Benchmark | GPT-5.5 | Hy4 flash | 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 | Not measured | Only GPT-5.5 measured |
| LMArena Text: medicine and healthcare | 1465 | Not measured | Only GPT-5.5 measured |
| LMArena Text: legal and government | 1492 | Not measured | Only GPT-5.5 measured |
| LMArena Text: business and finance | 1478 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 9 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index32
| Benchmark | GPT-5.5 | Hy4 flash | Leader |
|---|---|---|---|
| Design Arena: websites | 1243 | 1227 | GPT-5.5 |
| Design Arena: UI components | 1244 | 1262 | Hy4 flash |
| Design Arena: game dev | 1265 | 1202 | GPT-5.5 |
| Design Arena: data viz | 1257 | Not measured | Only GPT-5.5 measured |
| Design Arena: 3D | 1162 | 1209 | Hy4 flash |
| Design Arena: all code categories | 1242 | 1225 | GPT-5.5 |
| Design Arena: SVG | 1219 | Not measured | Only GPT-5.5 measured |
| Design Arena: ASCII art | 1237 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 1 board here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index39 boards, none shared
| Benchmark | GPT-5.5 | Hy4 flash | Leader |
|---|---|---|---|
| Vals Index | Not measured | 50.8% | Only Hy4 flash measured |
| Public Benefits Bench (Vals AI) | 60.9% | 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% | Not measured | Only GPT-5.5 measured |
| Long context reasoning (AA-LCR) | 84.3% | Not measured | Only GPT-5.5 measured |
| GDPval-AA | 1357 | Not measured | Only GPT-5.5 measured |
| 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 | Not measured | Only GPT-5.5 measured |
| LMArena Text: instruction following | 1478 | Not measured | Only GPT-5.5 measured |
| LMArena Text: creative writing | 1453 | Not measured | Only GPT-5.5 measured |
| LMArena Vision | 1296 | Not measured | Only GPT-5.5 measured |
| LMArena Text: English | 1470 | Not measured | Only GPT-5.5 measured |
| LMArena Text: non-English | 1466 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Chinese | 1534 | Not measured | Only GPT-5.5 measured |
| LMArena Text: German | 1481 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Japanese | 1502 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Korean | 1452 | Not measured | Only GPT-5.5 measured |
| LMArena Text: French | 1484 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Spanish | 1468 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Russian | 1474 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Polish | 1464 | Not measured | Only GPT-5.5 measured |
| LMArena Text: multi-turn | 1477 | Not measured | Only GPT-5.5 measured |
| LMArena Text: longer queries | 1483 | Not measured | Only GPT-5.5 measured |
| LMArena Text: writing and language | 1465 | Not measured | Only GPT-5.5 measured |
| LMArena Text: entertainment and media | 1448 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: English | 1294 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: Chinese | 1358 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: reading text | 1306 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: humor | 1297 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: recognizing things | 1229 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: creative writing | 1305 | Not measured | Only GPT-5.5 measured |
| LMArena Document | 1484 | Not measured | Only GPT-5.5 measured |
| EQ-Bench 4 | 1295 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 10 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5.5 | Hy4 flash | 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 flash costs less
Repository review
50K input + 3K output tokens
Hy4 flash costs less
Agent loop
200K input + 10K output tokens
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.
| Specification | OpenAI GPT-5.5 | Tencent Hy4 flash |
|---|---|---|
| Maker | OpenAI | Tencent |
| Weights | No public weights listed | Open weights |
| Released | April 24, 2026 | November 16, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $5.00 | $0.035 |
| Output $/Mtok | $30.00 | $0.1405 |
| Coding index | 67.9 | 44.5 |
| Independent measurements behind it | 4 | 2 |
| Cost per solved task | $1.65 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-5.5 better than Hy4 flash for coding?
Pick GPT-5.5 for coding: it scores 67.9 on the Vibe Coding Index against 44.5 for Hy4 flash. Pick Hy4 flash for output tokens 213.5x cheaper ($0.1405 vs $30.00 per million) and work measured by Vibe Code Bench (Vals AI) or Tax Agent Bench (Vals AI), where it leads.
Which is cheaper, GPT-5.5 or Hy4 flash?
GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens; Hy4 flash costs $0.035 and $0.1405. Cost per solved task is known for GPT-5.5 only.
Which has the bigger context window, GPT-5.5 or Hy4 flash?
GPT-5.5, with 1.05M tokens against 1.05M.
How many benchmarks is the GPT-5.5 vs Hy4 flash comparison based on?
9 benchmarks have a result for both models as of November 22, 2026. GPT-5.5'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.