Qwen3.5-Flash vs Hy4 preview
Hy4 preview leads Qwen3.5-Flash on coding, 44.3 to 41.1 on the Vibe Coding Index, and Qwen3.5-Flash costs 9.6x less per output token ($0.26 vs $2.501 per million). Updated October 22, 2026; 6 benchmarks have a result for both.
Qwen3.5-Flash or Hy4 preview: which to pick
Pick Hy4 preview for coding: it scores 44.3 on the Vibe Coding Index against 41.1 for Qwen3.5-Flash.
Pick Qwen3.5-Flash for output tokens 9.6x cheaper ($0.26 vs $2.501 per million) and work measured by LegalBench (Vals AI), where it leads.
- Shared results
- 6
- Qwen3.5-Flash only
- 49
- Hy4 preview only
- 18
- 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
Hy4 preview
Hy4 preview scores 44.3 against 41.1.
2 and 3 benchmarks behind the scores
Reasoning
GPQA Diamond
No call
Only Qwen3.5-Flash has a GPQA Diamond result, 82.8%.
Directional only
Math
AIME
No call
Only Qwen3.5-Flash has a AIME result, 92.5%.
Directional only
Knowledge
MMLU Pro
No call
Only Qwen3.5-Flash has a MMLU Pro result, 84.1%.
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 1M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen3.5-Flash
Qwen3.5-Flash $0.0002, Hy4 preview $0.0021 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Qwen3.5-Flash has the number, $0.40.
Directional only
At a glance
- Coding index
- Qwen3.5-Flash 41.1 from 2 benchmarks, Hy4 preview 44.3 from 3 benchmarks.
- Widest category
- Coding, 56.2 points apart on average over 3 shared benchmarks.
- Widest single gap
- LMArena WebDev: React: Qwen3.5-Flash 1229 against Hy4 preview 1653, 61.8 points apart on a 0 to 100 scale.
- Output price
- Qwen3.5-Flash $0.26 against Hy4 preview $2.501 per million output tokens, 9.6x apart.
- Context window
- Qwen3.5-Flash takes 1M 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 | Qwen3.5-Flash | Bars | Hy4 preview | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Knowledge | 1 | 1 | ||
| General | 0 | 1 | ||
| All shared boards | 1 | 5 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index03
| Benchmark | Qwen3.5-Flash | Hy4 preview | Leader |
|---|---|---|---|
| LMArena WebDev | 1244 | 1631 | Hy4 preview |
| LiveCodeBench (Vals AI) | 83.3% | Not measured | Only Qwen3.5-Flash measured |
| SWE-bench Verified (Vals AI) | 64.4% | Not measured | Only Qwen3.5-Flash measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 8.1% | Only Hy4 preview measured |
| Vibe Code Bench (Vals AI) | Not measured | 77.5% | Only Hy4 preview measured |
| 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 |
| ALE-Bench | 265.9 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: coding | 1412 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: software and IT | 1416 | Not measured | Only Qwen3.5-Flash measured |
| LMArena WebDev: React | 1229 | 1653 | Hy4 preview |
| LMArena WebDev: HTML | 1291 | 1620 | Hy4 preview |
| Neither model has a result on the other 28 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | Qwen3.5-Flash | Hy4 preview | Leader |
|---|---|---|---|
| Epoch Capabilities Index | 143.5 | Not measured | Only Qwen3.5-Flash measured |
| Neither model has a result on the other 1 board here | |||
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Qwen3.5-Flash | 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 |
| Neither model has a result on the other 34 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Qwen3.5-Flash | Hy4 preview | Leader |
|---|---|---|---|
| GPQA Diamond | 82.8% | Not measured | Only Qwen3.5-Flash measured |
| MMMU Pro (Vals AI) | 81.9% | Not measured | Only Qwen3.5-Flash measured |
| GPQA Diamond (Epoch AI run) | 82.3% | Not measured | Only Qwen3.5-Flash measured |
| Chess puzzles | 14.0% | Not measured | Only Qwen3.5-Flash measured |
| Mystery game puzzles | 20.0% | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: hard prompts | 1403 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: expert | 1407 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: hard prompts in English | 1409 | Not measured | Only Qwen3.5-Flash measured |
| Neither model has a result on the other 15 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Qwen3.5-Flash | Hy4 preview | Leader |
|---|---|---|---|
| AIME | 92.5% | Not measured | Only Qwen3.5-Flash measured |
| OTIS Mock AIME | 84.4% | Not measured | Only Qwen3.5-Flash measured |
| FrontierMath, tiers 1 to 3 | 18.2% | Not measured | Only Qwen3.5-Flash measured |
| ProofBench | Not measured | 75.0% | Only Hy4 preview measured |
| LMArena Text: math | 1407 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: mathematical work | 1413 | Not measured | Only Qwen3.5-Flash measured |
| Neither model has a result on the other 8 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | Qwen3.5-Flash | Hy4 preview | Leader |
|---|---|---|---|
| MMLU Pro | 84.1% | Not measured | Only Qwen3.5-Flash measured |
| LegalBench (Vals AI) | 84.3% | 83.8% | Qwen3.5-Flash |
| CaseLaw (Vals AI) | 56.0% | Not measured | Only Qwen3.5-Flash measured |
| CorpFin (Vals AI) | 63.6% | Not measured | Only Qwen3.5-Flash measured |
| TaxEval (Vals AI) | 72.2% | Not measured | Only Qwen3.5-Flash measured |
| MortgageTax (Vals AI) | 67.4% | Not measured | Only Qwen3.5-Flash measured |
| MedCode (Vals AI) | 33.0% | 43.3% | Hy4 preview |
| SimpleQA Verified | 20.3% | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: science | 1415 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: medicine and healthcare | 1407 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: legal and government | 1398 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: business and finance | 1394 | Not measured | Only Qwen3.5-Flash measured |
| Neither model has a result on the other 8 boards here | |||
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen3.5-Flash | Hy4 preview | Leader |
|---|---|---|---|
| Neither model has a result on 9 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen3.5-Flash | 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) | 42.5% | Not measured | Only Qwen3.5-Flash measured |
| MedScribe (Vals AI) | 70.6% | 83.6% | Hy4 preview |
| DTBench | 82.9% | Not measured | Only Qwen3.5-Flash measured |
| LMCA | 29.1% | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text | 1397 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: instruction following | 1374 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: creative writing | 1343 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: English | 1404 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: non-English | 1385 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: Chinese | 1445 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: German | 1390 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: Japanese | 1369 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: Korean | 1344 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: French | 1413 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: Spanish | 1400 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: Russian | 1380 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: Polish | 1376 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: multi-turn | 1393 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: longer queries | 1392 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: writing and language | 1369 | Not measured | Only Qwen3.5-Flash measured |
| LMArena Text: entertainment and media | 1347 | Not measured | Only Qwen3.5-Flash measured |
| Neither model has a result on the other 25 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen3.5-Flash | 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
Qwen3.5-Flash costs less
Repository review
50K input + 3K output tokens
Qwen3.5-Flash costs less
Agent loop
200K input + 10K output tokens
Qwen3.5-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 | Qwen3.5-Flash | Tencent Hy4 preview |
|---|---|---|
| Maker | Qwen | Tencent |
| Weights | No public weights listed | Open weights |
| Released | February 25, 2026 | August 28, 2026 |
| Context window | 1M | 1.05M |
| Input $/Mtok | $0.065 | $0.834 |
| Output $/Mtok | $0.26 | $2.501 |
| Coding index | 41.1 | 44.3 |
| Independent measurements behind it | 2 | 3 |
| Cost per solved task | $0.40 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Qwen3.5-Flash better than Hy4 preview for coding?
Pick Hy4 preview for coding: it scores 44.3 on the Vibe Coding Index against 41.1 for Qwen3.5-Flash. Pick Qwen3.5-Flash for output tokens 9.6x cheaper ($0.26 vs $2.501 per million) and work measured by LegalBench (Vals AI), where it leads.
Which is cheaper, Qwen3.5-Flash or Hy4 preview?
Qwen3.5-Flash costs $0.065 per million input tokens and $0.26 per million output tokens; Hy4 preview costs $0.834 and $2.501. Cost per solved task is known for Qwen3.5-Flash only.
Which has the bigger context window, Qwen3.5-Flash or Hy4 preview?
Hy4 preview, with 1.05M tokens against 1M.
How many benchmarks is the Qwen3.5-Flash vs Hy4 preview comparison based on?
6 benchmarks have a result for both models as of October 22, 2026. Qwen3.5-Flash's coding index rests on 2 independent benchmarks and Hy4 preview's on 3. Where only one model has a result, the page says so instead of naming a winner.