Qwen3 30B A3B Instruct 2507 vs Qwen3 32B
Qwen3 30B A3B Instruct 2507 leads Qwen3 32B on the Artificial Analysis Intelligence Index, 7.5 to 7.3, and Qwen3 30B A3B Instruct 2507 costs 1.3x less per output token ($0.30 vs $0.40 per million). Updated October 22, 2026; 40 benchmarks have a result for both.
Qwen3 30B A3B Instruct 2507 or Qwen3 32B: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick Qwen3 30B A3B Instruct 2507 for output tokens 1.3x cheaper ($0.30 vs $0.40 per million) and a larger context window (262K against 131K tokens).
- Shared results
- 40
- Qwen3 30B A3B Instruct 2507 only
- 5
- Qwen3 32B only
- 5
- 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
Qwen3 30B A3B Instruct 2507
Qwen3 30B A3B Instruct 2507 leads, 7.5 against 7.3.
Same board, both measured
Long documents
Listed context window
Qwen3 30B A3B Instruct 2507
Qwen3 30B A3B Instruct 2507 takes 262K tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen3 30B A3B Instruct 2507
Qwen3 30B A3B Instruct 2507 $0.0002, Qwen3 32B $0.0003 per turn.
List prices
At a glance
- Coding index
- Qwen3 30B A3B Instruct 2507 not benchmarked, Qwen3 32B not benchmarked.
- Widest category
- General, 11.0 points apart on average over 16 shared benchmarks.
- Widest single gap
- Long context reasoning (AA-LCR): Qwen3 30B A3B Instruct 2507 61.3% against Qwen3 32B 0.0%, 61.3 points apart on a 0 to 100 scale.
- Output price
- Qwen3 30B A3B Instruct 2507 $0.30 against Qwen3 32B $0.40 per million output tokens, 1.3x apart.
- Context window
- Qwen3 30B A3B Instruct 2507 takes 262K tokens, Qwen3 32B 131K.
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 30B A3B Instruct 2507 | Bars | Qwen3 32B | Ties |
|---|---|---|---|---|
| Coding | 4 | 1 | ||
| Intelligence | 1 | 0 | ||
| Agentic work | 0 | 2 | ||
| Reasoning | 5 | 1 | ||
| Math | 0 | 5 | ||
| Knowledge | 5 | 0 | ||
| General | 15 | 1 | ||
| All shared boards | 30 | 10 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index41
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| Aider polyglot | Not measured | 40.0% | Only Qwen3 32B measured |
| LiveCodeBench (Artificial Analysis run) | 70.7% | 54.6% | Qwen3 30B A3B Instruct 2507 |
| SciCode | 33.0% | 36.0% | Qwen3 32B |
| Terminal-Bench Hard (Artificial Analysis run) | 6.1% | 3.0% | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: coding | 1416 | 1357 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: software and IT | 1417 | 1366 | Qwen3 30B A3B Instruct 2507 |
| Neither model has a result on the other 33 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| AA Intelligence Index | 7.5 | 7.3 | Qwen3 30B A3B Instruct 2507 |
| Epoch Capabilities Index | Not measured | 138.4 | Only Qwen3 32B measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 28.1% | 29.8% | Qwen3 32B |
| Berkeley Function Calling Leaderboard (BFCL) | 41.4% | 48.7% | Qwen3 32B |
| BrowseComp-Plus | Not measured | 10.7% | Only Qwen3 32B measured |
| Neither model has a result on the other 38 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index51
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 70.7% | 66.8% | Qwen3 30B A3B Instruct 2507 |
| Humanity's Last Exam (Artificial Analysis run) | 10.3% | 7.4% | Qwen3 30B A3B Instruct 2507 |
| GPQA Diamond (Epoch AI run) | 55.6% | 65.7% | Qwen3 32B |
| LMArena Text: hard prompts | 1398 | 1334 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: expert | 1397 | 1362 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: hard prompts in English | 1406 | 1350 | Qwen3 30B A3B Instruct 2507 |
| Kagi LLM Benchmark | Not measured | 48.7% | Only Qwen3 32B measured |
| Neither model has a result on the other 15 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| AA Math Index | 66.3 | 73.0 | Qwen3 32B |
| AIME 2025 (Artificial Analysis run) | 66.3% | 73.0% | Qwen3 32B |
| OTIS Mock AIME | 62.2% | 66.9% | Qwen3 32B |
| LMArena Text: math | 1394 | 1399 | Qwen3 32B |
| LMArena Text: mathematical work | 1411 | 1420 | Qwen3 32B |
| Neither model has a result on the other 8 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 80.5% | 79.8% | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: science | 1393 | 1372 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: medicine and healthcare | 1392 | 1368 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: legal and government | 1387 | 1324 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: business and finance | 1389 | 1318 | Qwen3 30B A3B Instruct 2507 |
| Neither model has a result on the other 13 boards here | |||
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| Neither model has a result on 9 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index151
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | Leader |
|---|---|---|---|
| IFBench | 50.7% | 36.3% | Qwen3 30B A3B Instruct 2507 |
| Long context reasoning (AA-LCR) | 61.3% | 0.0% | Qwen3 30B A3B Instruct 2507 |
| DTBench | 67.2% | 67.5% | Qwen3 32B |
| LMCA | 22.4% | 17.3% | Qwen3 30B A3B Instruct 2507 |
| Fiction.LiveBench | Not measured | 74.2% | Only Qwen3 32B measured |
| LMArena Text | 1384 | 1340 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: instruction following | 1363 | 1305 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: creative writing | 1317 | 1297 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: English | 1391 | 1359 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: non-English | 1372 | 1317 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: Chinese | 1432 | 1357 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: German | 1379 | 1341 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: Japanese | 1337 | Not measured | Only Qwen3 30B A3B Instruct 2507 measured |
| LMArena Text: Korean | 1330 | Not measured | Only Qwen3 30B A3B Instruct 2507 measured |
| LMArena Text: French | 1419 | Not measured | Only Qwen3 30B A3B Instruct 2507 measured |
| LMArena Text: Spanish | 1404 | Not measured | Only Qwen3 30B A3B Instruct 2507 measured |
| LMArena Text: Russian | 1369 | 1311 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: Polish | 1353 | Not measured | Only Qwen3 30B A3B Instruct 2507 measured |
| LMArena Text: multi-turn | 1378 | 1331 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: longer queries | 1378 | 1327 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: writing and language | 1336 | 1298 | Qwen3 30B A3B Instruct 2507 |
| LMArena Text: entertainment and media | 1334 | 1279 | Qwen3 30B A3B Instruct 2507 |
| Neither model has a result on the other 24 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen3 30B A3B Instruct 2507 | Qwen3 32B | 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 30B A3B Instruct 2507 costs less
Repository review
50K input + 3K output tokens
Qwen3 30B A3B Instruct 2507 costs less
Agent loop
200K input + 10K output tokens
Qwen3 30B A3B Instruct 2507 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 30B A3B Instruct 2507 | Qwen3 32B |
|---|---|---|
| Maker | Qwen | Qwen |
| Weights | Open weights | Open weights |
| Released | July 29, 2025 | April 28, 2025 |
| Context window | 262K | 131K |
| Input $/Mtok | $0.09 | $0.14 |
| Output $/Mtok | $0.30 | $0.40 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Qwen3 30B A3B Instruct 2507 better than Qwen3 32B for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Qwen3 30B A3B Instruct 2507 for output tokens 1.3x cheaper ($0.30 vs $0.40 per million) and a larger context window (262K against 131K tokens).
Which is cheaper, Qwen3 30B A3B Instruct 2507 or Qwen3 32B?
Qwen3 30B A3B Instruct 2507 costs $0.09 per million input tokens and $0.30 per million output tokens; Qwen3 32B costs $0.14 and $0.40. Cost per solved task is known for neither of them.
Which has the bigger context window, Qwen3 30B A3B Instruct 2507 or Qwen3 32B?
Qwen3 30B A3B Instruct 2507, with 262K tokens against 131K.
How many benchmarks is the Qwen3 30B A3B Instruct 2507 vs Qwen3 32B comparison based on?
40 benchmarks have a result for both models as of October 22, 2026. Qwen3 30B A3B Instruct 2507's coding index rests on 0 independent benchmarks and Qwen3 32B's on 0. Where only one model has a result, the page says so instead of naming a winner.