Qwen3.8 27B vs Hy4 preview
Updated October 21, 2026. 18 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 61.6 for Hy4 preview.
Pick Hy4 preview when the context window is larger, 1.05M against 1M tokens, or when what matters is LMArena WebDev or Vibe Code Bench (Vals AI), where it leads.
- Shared results
- 18
- Qwen3.8 27B only
- 67
- Hy4 preview only
- 6
- Like-for-like categories
- 5 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
Qwen3.8 27B
Qwen3.8 27B scores 73.6 against 61.6.
2 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
No call
Only Qwen3.8 27B has a AA Intelligence Index result, 33.7.
Directional only
Agentic work
AA Agentic Index
No call
Only Qwen3.8 27B has a AA Agentic Index result, 45.8.
Directional only
Reasoning
GPQA Diamond
No call
Only Qwen3.8 27B has a GPQA Diamond result, 88.9%.
Directional only
Knowledge
MMLU Pro
No call
Only Qwen3.8 27B has a MMLU Pro result, 84.3%.
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
Qwen3.8 27B
Qwen3.8 27B $0.0007, Hy4 preview $0.0021 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Qwen3.8 27B has the number, $1.26.
Directional only
At a glance
- Coding index
- Qwen3.8 27B 73.6 from 2 benchmarks, Hy4 preview 61.6 from 1 benchmark.
- Widest category
- Math, 59.0 points apart on average over 1 shared benchmark.
- Widest single gap
- ProofBench: Qwen3.8 27B 16.0% against Hy4 preview 75.0%, 59.0 points apart on a 0 to 100 scale.
- Output price
- Qwen3.8 27B $1.35 against Hy4 preview $2.501 per million output tokens, 1.9x apart.
- Context window
- Qwen3.8 27B 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 | Qwen3.8 27B | Bars | Hy4 preview | Ties |
|---|---|---|---|---|
| Coding | 2 | 6 | 1 tie | |
| Agentic work | 1 | 4 | ||
| Math | 0 | 1 | ||
| Knowledge | 0 | 2 | ||
| General | 1 | 0 | ||
| All shared boards | 4 | 13 | 1 tie |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index26
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| LiveBench coding | 86.8% | Not measured | Only Qwen3.8 27B measured |
| LMArena WebDev | 1593 | 1631 | Hy4 preview |
| AA Coding Index | 68.1 | Not measured | Only Qwen3.8 27B measured |
| LiveCodeBench (Vals AI) | 84.0% | Not measured | Only Qwen3.8 27B measured |
| SWE-bench Verified (Vals AI) | 86.0% | Not measured | Only Qwen3.8 27B measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 8.1% | Only Hy4 preview measured |
| Vibe Code Bench (Vals AI) | 64.8% | 77.5% | Hy4 preview |
| Code migration (Vals AI) | 14.2% | 47.4% | Hy4 preview |
| ProgramBench (Vals AI) | 0.0% | 0.0% | Tie |
| IOI (Vals AI) | 39.1% | 59.3% | Hy4 preview |
| Terminal-Bench 2.1 (Vals AI) | 58.4% | 55.1% | Qwen3.8 27B |
| SkillsBench (Vals AI) | 38.1% | Not measured | Only Qwen3.8 27B measured |
| CyberBench (Vals AI) | 67.8% | 65.4% | Qwen3.8 27B |
| SciCode | 46.6% | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: coding | 1483 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: software and IT | 1478 | Not measured | Only Qwen3.8 27B measured |
| LMArena WebDev: React | 1602 | 1653 | Hy4 preview |
| LMArena WebDev: HTML | 1570 | 1620 | Hy4 preview |
| LMArena WebDev: from an image | 1577 | Not measured | Only Qwen3.8 27B measured |
| Neither model has a result on the other 19 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| AA Intelligence Index | 33.7 | Not measured | Only Qwen3.8 27B measured |
| Epoch Capabilities Index | 151.7 | Not measured | Only Qwen3.8 27B measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index14
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| AA Agentic Index | 45.8 | Not measured | Only Qwen3.8 27B measured |
| LiveBench research | 59.4% | Not measured | Only Qwen3.8 27B measured |
| LiveBench browser | 32.0% | Not measured | Only Qwen3.8 27B measured |
| LiveBench long agentic | 27.9% | Not measured | Only Qwen3.8 27B measured |
| SREBench (Vals AI) | Not measured | 2.3% | Only Hy4 preview measured |
| Finance Agent (Vals AI) | 48.5% | 55.1% | Hy4 preview |
| Tax Agent Bench (Vals AI) | 30.9% | 32.6% | Hy4 preview |
| Legal Research Bench (Vals AI) | 36.1% | 45.2% | Hy4 preview |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 8.8% | Only Hy4 preview measured |
| Excel Modeling Benchmark (Vals AI) | 59.7% | 58.1% | Qwen3.8 27B |
| BioMysteryBench (Vals AI) | Not measured | 69.3% | Only Hy4 preview measured |
| Terminal-Bench-Science (Vals AI) | 0.0% | 1.4% | Hy4 preview |
| APEX-Agents | 47.5% | Not measured | Only Qwen3.8 27B measured |
| Neither model has a result on the other 26 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index12 boards, none shared
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| GPQA Diamond | 88.9% | Not measured | Only Qwen3.8 27B measured |
| LiveBench reasoning | 80.0% | Not measured | Only Qwen3.8 27B measured |
| MMMU Pro (Vals AI) | 83.9% | Not measured | Only Qwen3.8 27B measured |
| GPQA Diamond (Artificial Analysis run) | 90.5% | Not measured | Only Qwen3.8 27B measured |
| Humanity's Last Exam (Artificial Analysis run) | 33.9% | Not measured | Only Qwen3.8 27B measured |
| ARC-AGI | 87.5% | Not measured | Only Qwen3.8 27B measured |
| ARC-AGI-2 | 42.4% | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: hard prompts | 1459 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: expert | 1479 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: hard prompts in English | 1461 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision: diagrams | 1285 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision: homework | 1295 | Not measured | Only Qwen3.8 27B measured |
| Neither model has a result on the other 9 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| ProofBench | 16.0% | 75.0% | Hy4 preview |
| LMArena Text: math | 1455 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: mathematical work | 1458 | Not measured | Only Qwen3.8 27B measured |
| Neither model has a result on the other 10 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| MMLU Pro | 84.3% | Not measured | Only Qwen3.8 27B measured |
| LegalBench (Vals AI) | 82.4% | 83.8% | Hy4 preview |
| TaxEval (Vals AI) | 70.8% | Not measured | Only Qwen3.8 27B measured |
| MortgageTax (Vals AI) | 64.9% | Not measured | Only Qwen3.8 27B measured |
| MedCode (Vals AI) | 28.7% | 43.3% | Hy4 preview |
| LMArena Text: science | 1467 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: medicine and healthcare | 1437 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: legal and government | 1447 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: business and finance | 1438 | Not measured | Only Qwen3.8 27B measured |
| Neither model has a result on the other 8 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| LiveBench design | 75.9% | Not measured | Only Qwen3.8 27B measured |
| Neither model has a result on the other 8 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Qwen3.8 27B | Hy4 preview | Leader |
|---|---|---|---|
| Vals Index | Not measured | 53.2% | Only Hy4 preview measured |
| LiveBench data analysis | 76.6% | Not measured | Only Qwen3.8 27B measured |
| Public Benefits Bench (Vals AI) | Not measured | 68.6% | Only Hy4 preview measured |
| SAGE (Vals AI) | 52.4% | Not measured | Only Qwen3.8 27B measured |
| MedScribe (Vals AI) | 83.8% | 83.6% | Qwen3.8 27B |
| Long context reasoning (AA-LCR) | 82.0% | Not measured | Only Qwen3.8 27B measured |
| DTBench | 88.0% | Not measured | Only Qwen3.8 27B measured |
| LMCA | 41.4% | Not measured | Only Qwen3.8 27B measured |
| Surface Evolver Bench | 45.0% | Not measured | Only Qwen3.8 27B measured |
| LMArena Text | 1440 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: instruction following | 1436 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: creative writing | 1383 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision | 1269 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: English | 1448 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: non-English | 1428 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: Chinese | 1501 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: German | 1443 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: Japanese | 1387 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: Korean | 1384 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: French | 1465 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: Spanish | 1449 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: Russian | 1415 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: Polish | 1439 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: multi-turn | 1440 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: longer queries | 1448 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: writing and language | 1405 | Not measured | Only Qwen3.8 27B measured |
| LMArena Text: entertainment and media | 1393 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision: English | 1267 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision: Chinese | 1326 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision: reading text | 1281 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision: humor | 1235 | Not measured | Only Qwen3.8 27B measured |
| LMArena Vision: creative writing | 1237 | Not measured | Only Qwen3.8 27B measured |
| Neither model has a result on the other 14 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen3.8 27B | 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
Qwen3.8 27B costs less
Repository review
50K input + 3K output tokens
Qwen3.8 27B costs less
Agent loop
200K input + 10K output tokens
Qwen3.8 27B 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 | Qwen3.8 27B | Tencent Hy4 preview |
|---|---|---|
| Maker | Qwen | Tencent |
| Weights | Open weights | Open weights |
| Released | August 14, 2026 | August 28, 2026 |
| Context window | 1M | 1.05M |
| Input $/Mtok | $0.04 | $0.834 |
| Output $/Mtok | $1.35 | $2.501 |
| Coding index | 73.6 | 61.6 |
| Independent measurements behind it | 2 | 1 |
| Cost per solved task | $1.26 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Qwen3.8 27B better than Hy4 preview for coding?
Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 61.6 for Hy4 preview. Pick Hy4 preview when the context window is larger, 1.05M against 1M tokens, or when what matters is LMArena WebDev or Vibe Code Bench (Vals AI), where it leads.
Which is cheaper, Qwen3.8 27B or Hy4 preview?
Qwen3.8 27B costs $0.04 per million input tokens and $1.35 per million output tokens; Hy4 preview costs $0.834 and $2.501. Cost per solved task is known for Qwen3.8 27B only.
Which has the bigger context window, Qwen3.8 27B or Hy4 preview?
Hy4 preview, with 1.05M tokens against 1M.
How many benchmarks is the Qwen3.8 27B vs Hy4 preview comparison based on?
18 benchmarks have a result for both models as of October 21, 2026. Qwen3.8 27B's coding index rests on 2 independent benchmarks and Hy4 preview's on 1. Where only one model has a result, the page says so instead of naming a winner.