Qwen3.8 Max (0902) vs Qwen4 Max
Qwen4 Max leads Qwen3.8 Max (0902) on coding, 74.0 to 67.2 on the Vibe Coding Index, and Qwen4 Max costs 1.4x less per output token ($4.275 vs $6.00 per million). Updated November 22, 2026; 26 benchmarks have a result for both.
Qwen3.8 Max (0902) or Qwen4 Max: which to pick
Pick Qwen4 Max for coding: it scores 74.0 on the Vibe Coding Index against 67.2 for Qwen3.8 Max (0902), though on 4 benchmarks against 5.
Pick Qwen3.8 Max (0902) for work measured by Tax Agent Bench (Vals AI) or Legal Research Bench (Vals AI), where it leads.
- Shared results
- 26
- Qwen3.8 Max (0902) only
- 101
- Qwen4 Max only
- 0
- Like-for-like categories
- 7 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
Qwen4 Max
Qwen4 Max scores 74.0 against 67.2.
5 and 4 benchmarks behind the scores
Intelligence
AA Intelligence Index
Qwen4 Max
Qwen4 Max leads, 42.9 against 33.7.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Qwen3.8 Max (0902) has a GPQA Diamond result, 93.7%.
Directional only
Knowledge
MMLU Pro
No call
Only Qwen3.8 Max (0902) has a MMLU Pro result, 88.6%.
Directional only
Web design
Design Arena: websites
No call
Only Qwen3.8 Max (0902) has a Design Arena: websites result, 1264.
Directional only
General
Vals Index
Qwen4 Max
Qwen4 Max leads, 63.2% against 59.1%.
Same board, both measured
Long documents
Listed context window
No call
Both take 1M tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen4 Max
Qwen3.8 Max (0902) $0.0050, Qwen4 Max $0.0036 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Qwen3.8 Max (0902) has the number, $1.32.
Directional only
At a glance
- Coding index
- Qwen3.8 Max (0902) 67.2 from 5 benchmarks, Qwen4 Max 74.0 from 4 benchmarks.
- Widest category
- Agentic work, 14.1 points apart on average over 6 shared benchmarks.
- Widest single gap
- LiveBench browser: Qwen3.8 Max (0902) 49.4% against Qwen4 Max 80.7%, 31.3 points apart on a 0 to 100 scale.
- Output price
- Qwen3.8 Max (0902) $6.00 against Qwen4 Max $4.275 per million output tokens, 1.4x apart.
- Context window
- Both take 1M tokens.
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 Max (0902) | Bars | Qwen4 Max | Ties |
|---|---|---|---|---|
| Coding | 0 | 11 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 2 | 4 | ||
| Reasoning | 0 | 3 | ||
| Knowledge | 0 | 1 | ||
| Web design | 0 | 1 | ||
| General | 0 | 3 | ||
| All shared boards | 2 | 24 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index011
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| LiveBench coding | 60.9% | 69.8% | Qwen4 Max |
| LMArena WebDev | 1674 | 1751 | Qwen4 Max |
| LiveCodeBench (Vals AI) | 87.8% | Not measured | Only Qwen3.8 Max (0902) measured |
| SWE-bench Verified (Vals AI) | 85.6% | Not measured | Only Qwen3.8 Max (0902) measured |
| Terminal-Bench 4 (Vals AI) | 38.9% | 48.5% | Qwen4 Max |
| Vibe Code Bench (Vals AI) | 77.1% | 86.7% | Qwen4 Max |
| Code migration (Vals AI) | 24.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| IOI (Vals AI) | 68.9% | Not measured | Only Qwen3.8 Max (0902) measured |
| Terminal-Bench 2.1 (Vals AI) | 67.4% | Not measured | Only Qwen3.8 Max (0902) measured |
| SWE-bench Multimodal (Vals AI) | 42.5% | Not measured | Only Qwen3.8 Max (0902) measured |
| Vibe Code Bench 1-100 (Vals AI) | 24.8% | Not measured | Only Qwen3.8 Max (0902) measured |
| SkillsBench (Vals AI) | 42.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| CyberBench (Vals AI) | 28.6% | Not measured | Only Qwen3.8 Max (0902) measured |
| SciCode | 53.2% | Not measured | Only Qwen3.8 Max (0902) measured |
| Terminal-Bench 4.0 (Artificial Analysis run) | 38.9% | 48.5% | Qwen4 Max |
| FrontierCode 2.0 | 23.4% | 26.5% | Qwen4 Max |
| FrontierSWE | 17.8% | Not measured | Only Qwen3.8 Max (0902) measured |
| MirrorCode | 52.2% | Not measured | Only Qwen3.8 Max (0902) measured |
| ALE-Bench | 1679.1 | 1831.0 | Qwen4 Max |
| LMArena Text: coding | 1503 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: software and IT | 1502 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena WebDev: React | 1696 | 1785 | Qwen4 Max |
| LMArena WebDev: HTML | 1653 | 1752 | Qwen4 Max |
| LMArena WebDev: from an image | 1639 | 1675 | Qwen4 Max |
| GameBench 2 | 69.7 | 82.4 | Qwen4 Max |
| Terminal-Bench 4.0 | 27.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| Neither model has a result on the other 25 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| AA Intelligence Index | 33.7 | 42.9 | Qwen4 Max |
| Epoch Capabilities Index | 157.6 | Not measured | Only Qwen3.8 Max (0902) measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index24
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| Design Arena: agents building web apps | 1270 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: agents building full-stack apps | 1282 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: agents building mobile apps | 1224 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: agents building Android apps | 1191 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: agents building Godot games | 1210 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: PowerPoint slides | 1243 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: spreadsheets | 1004 | Not measured | Only Qwen3.8 Max (0902) measured |
| LiveBench research | 72.8% | 84.7% | Qwen4 Max |
| LiveBench browser | 49.4% | 80.7% | Qwen4 Max |
| LiveBench long agentic | 26.6% | 46.6% | Qwen4 Max |
| Finance Agent (Vals AI) | 57.9% | Not measured | Only Qwen3.8 Max (0902) measured |
| Tax Agent Bench (Vals AI) | 49.7% | 40.4% | Qwen3.8 Max (0902) |
| Legal Research Bench (Vals AI) | 47.6% | 41.4% | Qwen3.8 Max (0902) |
| Harvey's Legal Agent Benchmark (Vals AI) | 15.4% | 21.3% | Qwen4 Max |
| Excel Modeling Benchmark (Vals AI) | 72.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| Terminal-Bench-Science (Vals AI) | 12.9% | Not measured | Only Qwen3.8 Max (0902) measured |
| Robo-bench (Vals AI) | 5.2% | Not measured | Only Qwen3.8 Max (0902) measured |
| APEX-Agents | 63.3% | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Search | 1261 | Not measured | Only Qwen3.8 Max (0902) measured |
| τ³-Banking | 55.2% | Not measured | Only Qwen3.8 Max (0902) measured |
| Neither model has a result on the other 24 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index03
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| GPQA Diamond | 93.7% | Not measured | Only Qwen3.8 Max (0902) measured |
| LiveBench reasoning | 88.2% | 92.2% | Qwen4 Max |
| MysteryMechanism (Vals AI) | 23.9% | Not measured | Only Qwen3.8 Max (0902) measured |
| MMMU Pro (Vals AI) | 88.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| GPQA Diamond (Artificial Analysis run) | 92.8% | Not measured | Only Qwen3.8 Max (0902) measured |
| Humanity's Last Exam (Artificial Analysis run) | 43.1% | Not measured | Only Qwen3.8 Max (0902) measured |
| HLE-Diamond (Artificial Analysis run) | 24.0% | 44.9% | Qwen4 Max |
| CritPt v2, hard subset (Artificial Analysis run) | 29.5% | 43.0% | Qwen4 Max |
| GPQA Diamond (Epoch AI run) | 92.7% | Not measured | Only Qwen3.8 Max (0902) measured |
| Chess puzzles | 46.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| Mystery game puzzles | 38.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: hard prompts | 1500 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: expert | 1510 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: hard prompts in English | 1505 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision: diagrams | 1320 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision: homework | 1333 | Not measured | Only Qwen3.8 Max (0902) measured |
| Neither model has a result on the other 10 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| OTIS Mock AIME | 100.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| FrontierMath, tiers 1 to 3 | 74.7% | Not measured | Only Qwen3.8 Max (0902) measured |
| FrontierMath, tier 4 | 46.3% | Not measured | Only Qwen3.8 Max (0902) measured |
| ProofBench | 58.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: math | 1505 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: mathematical work | 1495 | Not measured | Only Qwen3.8 Max (0902) measured |
| MathArena | 47.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| MMLU Pro | 88.6% | Not measured | Only Qwen3.8 Max (0902) measured |
| LegalBench (Vals AI) | 83.6% | Not measured | Only Qwen3.8 Max (0902) measured |
| CorpFin (Vals AI) | 65.8% | Not measured | Only Qwen3.8 Max (0902) measured |
| TaxEval (Vals AI) | 75.5% | Not measured | Only Qwen3.8 Max (0902) measured |
| MortgageTax (Vals AI) | 64.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| MedCode (Vals AI) | 40.7% | Not measured | Only Qwen3.8 Max (0902) measured |
| AA-Omniscience | 10.5 | 27.7 | Qwen4 Max |
| SimpleQA Verified | 47.3% | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: science | 1513 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: medicine and healthcare | 1512 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: legal and government | 1500 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: business and finance | 1481 | Not measured | Only Qwen3.8 Max (0902) measured |
| FACTS Benchmark Suite | 60.3% | Not measured | Only Qwen3.8 Max (0902) measured |
| Neither model has a result on the other 8 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| Design Arena: websites | 1264 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: UI components | 1325 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: game dev | 1295 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: data viz | 1279 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: 3D | 1306 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: all code categories | 1279 | Not measured | Only Qwen3.8 Max (0902) measured |
| Design Arena: ASCII art | 1184 | Not measured | Only Qwen3.8 Max (0902) measured |
| LiveBench design | 76.6% | 79.1% | Qwen4 Max |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index03
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | Leader |
|---|---|---|---|
| Vals Index | 59.1% | 63.2% | Qwen4 Max |
| LiveBench data analysis | 78.4% | 83.4% | Qwen4 Max |
| Public Benefits Bench (Vals AI) | 67.1% | Not measured | Only Qwen3.8 Max (0902) measured |
| SAGE (Vals AI) | 81.8% | Not measured | Only Qwen3.8 Max (0902) measured |
| MedScribe (Vals AI) | 85.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| Vals Multimodal Index | 65.4% | Not measured | Only Qwen3.8 Max (0902) measured |
| Long context reasoning (AA-LCR) | 80.3% | Not measured | Only Qwen3.8 Max (0902) measured |
| GDPval-AA | 1674 | 1809 | Qwen4 Max |
| DTBench | 92.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| LMCA | 46.2% | Not measured | Only Qwen3.8 Max (0902) measured |
| Furniture assembly | 20.0% | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text | 1485 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: instruction following | 1484 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: creative writing | 1484 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision | 1307 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: English | 1494 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: non-English | 1475 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: Chinese | 1544 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: German | 1474 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: Japanese | 1467 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: Korean | 1458 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: French | 1506 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: Spanish | 1492 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: Russian | 1487 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: Polish | 1482 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: multi-turn | 1492 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: longer queries | 1494 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: writing and language | 1480 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Text: entertainment and media | 1472 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision: English | 1306 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision: Chinese | 1354 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision: reading text | 1319 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision: humor | 1305 | Not measured | Only Qwen3.8 Max (0902) measured |
| LMArena Vision: creative writing | 1301 | Not measured | Only Qwen3.8 Max (0902) measured |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen3.8 Max (0902) | Qwen4 Max | 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
Qwen4 Max costs less
Repository review
50K input + 3K output tokens
Qwen4 Max costs less
Agent loop
200K input + 10K output tokens
Qwen4 Max 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 Max (0902) | Qwen4 Max |
|---|---|---|
| Maker | Qwen | Qwen |
| Weights | No public weights listed | No public weights listed |
| Released | September 3, 2026 | November 16, 2026 |
| Context window | 1M | 1M |
| Input $/Mtok | $2.00 | $1.425 |
| Output $/Mtok | $6.00 | $4.275 |
| Coding index | 67.2 | 74.0 |
| Independent measurements behind it | 5 | 4 |
| Cost per solved task | $1.32 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Qwen3.8 Max (0902) better than Qwen4 Max for coding?
Pick Qwen4 Max for coding: it scores 74.0 on the Vibe Coding Index against 67.2 for Qwen3.8 Max (0902), though on 4 benchmarks against 5. Pick Qwen3.8 Max (0902) for work measured by Tax Agent Bench (Vals AI) or Legal Research Bench (Vals AI), where it leads.
Which is cheaper, Qwen3.8 Max (0902) or Qwen4 Max?
Qwen3.8 Max (0902) costs $2.00 per million input tokens and $6.00 per million output tokens; Qwen4 Max costs $1.425 and $4.275. Cost per solved task is known for Qwen3.8 Max (0902) only.
Which has the bigger context window, Qwen3.8 Max (0902) or Qwen4 Max?
Neither: both take 1M tokens.
How many benchmarks is the Qwen3.8 Max (0902) vs Qwen4 Max comparison based on?
26 benchmarks have a result for both models as of November 22, 2026. Qwen3.8 Max (0902)'s coding index rests on 5 independent benchmarks and Qwen4 Max's on 4. Where only one model has a result, the page says so instead of naming a winner.