DeepSeek V3.2 Exp vs Qwen3 VL 235B A22B Instruct
DeepSeek V3.2 Exp leads Qwen3 VL 235B A22B Instruct on the Artificial Analysis Intelligence Index, 16.6 to 9.9, and DeepSeek V3.2 Exp costs 4.6x less per output token ($0.41 vs $1.90 per million). Updated October 22, 2026; 39 benchmarks have a result for both.
DeepSeek V3.2 Exp or Qwen3 VL 235B A22B Instruct: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick DeepSeek V3.2 Exp for output tokens 4.6x cheaper ($0.41 vs $1.90 per million) and work measured by LMArena Text: software and IT or τ²-Bench Telecom, where it leads.
- Shared results
- 39
- DeepSeek V3.2 Exp only
- 20
- Qwen3 VL 235B A22B Instruct only
- 10
- 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
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
DeepSeek V3.2 Exp
DeepSeek V3.2 Exp leads, 16.6 against 9.9.
Same board, both measured
Web design
Design Arena: websites
No call
Only DeepSeek V3.2 Exp has a Design Arena: websites result, 1178.
Directional only
Long documents
Listed context window
Qwen3 VL 235B A22B Instruct
Qwen3 VL 235B A22B Instruct takes 262K tokens against 164K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V3.2 Exp
DeepSeek V3.2 Exp $0.0005, Qwen3 VL 235B A22B Instruct $0.0012 per turn.
List prices
At a glance
- Coding index
- DeepSeek V3.2 Exp not benchmarked, Qwen3 VL 235B A22B Instruct not benchmarked.
- Widest category
- Agentic work, 20.2 points apart on average over 1 shared benchmark.
- Widest single gap
- τ²-Bench Telecom: DeepSeek V3.2 Exp 33.9% against Qwen3 VL 235B A22B Instruct 54.1%, 20.2 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V3.2 Exp $0.41 against Qwen3 VL 235B A22B Instruct $1.90 per million output tokens, 4.6x apart.
- Context window
- DeepSeek V3.2 Exp takes 164K tokens, Qwen3 VL 235B A22B Instruct 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 | DeepSeek V3.2 Exp | Bars | Qwen3 VL 235B A22B Instruct | Ties |
|---|---|---|---|---|
| Coding | 3 | 1 | ||
| Intelligence | 1 | 0 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 4 | 1 | ||
| Math | 1 | 3 | ||
| Knowledge | 2 | 3 | ||
| General | 10 | 9 | ||
| All shared boards | 21 | 18 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index31
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| Aider polyglot | 74.2% | Not measured | Only DeepSeek V3.2 Exp measured |
| LMArena WebDev | 1272 | Not measured | Only DeepSeek V3.2 Exp measured |
| LiveCodeBench (Artificial Analysis run) | 78.9% | 64.6% | DeepSeek V3.2 Exp |
| Terminal-Bench Hard (Artificial Analysis run) | 31.1% | 11.4% | DeepSeek V3.2 Exp |
| WeirdML | 39.5% | Not measured | Only DeepSeek V3.2 Exp measured |
| LMArena Text: coding | 1438 | 1438 | DeepSeek V3.2 Exp |
| LMArena Text: software and IT | 1438 | 1442 | Qwen3 VL 235B A22B Instruct |
| LMArena WebDev: HTML | 1281 | Not measured | Only DeepSeek V3.2 Exp measured |
| Neither model has a result on the other 31 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| AA Intelligence Index | 16.6 | 9.9 | DeepSeek V3.2 Exp |
| Epoch Capabilities Index | 144.9 | Not measured | Only DeepSeek V3.2 Exp measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 33.9% | 54.1% | Qwen3 VL 235B A22B Instruct |
| TheAgentCompany | 42.9% | Not measured | Only DeepSeek V3.2 Exp measured |
| Berkeley Function Calling Leaderboard (BFCL) | 56.7% | Not measured | Only DeepSeek V3.2 Exp measured |
| Neither model has a result on the other 38 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index41
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 79.7% | 77.2% | DeepSeek V3.2 Exp |
| Humanity's Last Exam (Artificial Analysis run) | 14.9% | 11.9% | DeepSeek V3.2 Exp |
| LMArena Text: hard prompts | 1429 | 1428 | DeepSeek V3.2 Exp |
| LMArena Text: expert | 1422 | 1442 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: hard prompts in English | 1444 | 1439 | DeepSeek V3.2 Exp |
| LMArena Vision: diagrams | Not measured | 1255 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Vision: homework | Not measured | 1277 | Only Qwen3 VL 235B A22B Instruct measured |
| Kagi LLM Benchmark | 52.2% | Not measured | Only DeepSeek V3.2 Exp measured |
| Neither model has a result on the other 14 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index13
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| AA Math Index | 87.7 | 88.3 | Qwen3 VL 235B A22B Instruct |
| AIME 2025 (Artificial Analysis run) | 87.7% | 88.3% | Qwen3 VL 235B A22B Instruct |
| LMArena Text: math | 1422 | 1425 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: mathematical work | 1433 | 1419 | DeepSeek V3.2 Exp |
| Neither model has a result on the other 9 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index23
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 85.0% | 83.6% | DeepSeek V3.2 Exp |
| LMArena Text: science | 1443 | 1444 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: medicine and healthcare | 1468 | 1431 | DeepSeek V3.2 Exp |
| LMArena Text: legal and government | 1419 | 1438 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: business and finance | 1412 | 1433 | Qwen3 VL 235B A22B Instruct |
| Neither model has a result on the other 13 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| Design Arena: websites | 1178 | Not measured | Only DeepSeek V3.2 Exp measured |
| Design Arena: UI components | 1164 | Not measured | Only DeepSeek V3.2 Exp measured |
| Design Arena: game dev | 1144 | Not measured | Only DeepSeek V3.2 Exp measured |
| Design Arena: data viz | 1152 | Not measured | Only DeepSeek V3.2 Exp measured |
| Design Arena: 3D | 1148 | Not measured | Only DeepSeek V3.2 Exp measured |
| Design Arena: all code categories | 1170 | Not measured | Only DeepSeek V3.2 Exp measured |
| Design Arena: SVG | 1031 | Not measured | Only DeepSeek V3.2 Exp measured |
| Neither model has a result on the other 2 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index109
| Benchmark | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | Leader |
|---|---|---|---|
| IFBench | 54.1% | 56.5% | Qwen3 VL 235B A22B Instruct |
| Long context reasoning (AA-LCR) | 72.3% | 63.7% | DeepSeek V3.2 Exp |
| DTBench | 87.7% | Not measured | Only DeepSeek V3.2 Exp measured |
| LMCA | 29.1% | Not measured | Only DeepSeek V3.2 Exp measured |
| Fiction.LiveBench | 83.3% | Not measured | Only DeepSeek V3.2 Exp measured |
| CL-bench | 13.2% | Not measured | Only DeepSeek V3.2 Exp measured |
| CL-bench Life | 9.5% | Not measured | Only DeepSeek V3.2 Exp measured |
| LMArena Text | 1425 | 1420 | DeepSeek V3.2 Exp |
| LMArena Text: instruction following | 1401 | 1406 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: creative writing | 1404 | 1366 | DeepSeek V3.2 Exp |
| LMArena Vision | Not measured | 1247 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Text: English | 1441 | 1429 | DeepSeek V3.2 Exp |
| LMArena Text: non-English | 1404 | 1405 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: Chinese | 1461 | 1463 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: German | 1440 | 1424 | DeepSeek V3.2 Exp |
| LMArena Text: Japanese | 1331 | 1385 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: Korean | 1372 | 1394 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: French | 1434 | 1452 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: Spanish | 1441 | 1428 | DeepSeek V3.2 Exp |
| LMArena Text: Russian | 1412 | 1408 | DeepSeek V3.2 Exp |
| LMArena Text: Polish | 1435 | 1419 | DeepSeek V3.2 Exp |
| LMArena Text: multi-turn | 1422 | 1428 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: longer queries | 1418 | 1419 | Qwen3 VL 235B A22B Instruct |
| LMArena Text: writing and language | 1407 | 1392 | DeepSeek V3.2 Exp |
| LMArena Text: entertainment and media | 1407 | 1368 | DeepSeek V3.2 Exp |
| LMArena Vision: English | Not measured | 1256 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Vision: Chinese | Not measured | 1283 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Vision: reading text | Not measured | 1253 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Vision: humor | Not measured | 1209 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Vision: captioning | Not measured | 1281 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Vision: recognizing things | Not measured | 1249 | Only Qwen3 VL 235B A22B Instruct measured |
| LMArena Vision: creative writing | Not measured | 1220 | Only Qwen3 VL 235B A22B Instruct 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 | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct | 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
DeepSeek V3.2 Exp costs less
Repository review
50K input + 3K output tokens
DeepSeek V3.2 Exp costs less
Agent loop
200K input + 10K output tokens
DeepSeek V3.2 Exp 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 | DeepSeek V3.2 Exp | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| Maker | DeepSeek | Qwen |
| Weights | Open weights | Open weights |
| Released | September 29, 2025 | September 23, 2025 |
| Context window | 164K | 262K |
| Input $/Mtok | $0.27 | $0.21 |
| Output $/Mtok | $0.41 | $1.90 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 1 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V3.2 Exp better than Qwen3 VL 235B A22B Instruct for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick DeepSeek V3.2 Exp for output tokens 4.6x cheaper ($0.41 vs $1.90 per million) and work measured by LMArena Text: software and IT or τ²-Bench Telecom, where it leads.
Which is cheaper, DeepSeek V3.2 Exp or Qwen3 VL 235B A22B Instruct?
DeepSeek V3.2 Exp costs $0.27 per million input tokens and $0.41 per million output tokens; Qwen3 VL 235B A22B Instruct costs $0.21 and $1.90. Cost per solved task is known for neither of them.
Which has the bigger context window, DeepSeek V3.2 Exp or Qwen3 VL 235B A22B Instruct?
Qwen3 VL 235B A22B Instruct, with 262K tokens against 164K.
How many benchmarks is the DeepSeek V3.2 Exp vs Qwen3 VL 235B A22B Instruct comparison based on?
39 benchmarks have a result for both models as of October 22, 2026. DeepSeek V3.2 Exp's coding index rests on 1 independent benchmark and Qwen3 VL 235B A22B Instruct's on 0. Where only one model has a result, the page says so instead of naming a winner.