Model comparison
Qwen3 235B-A22B vs Qwen3 32B
Qwen3 235B-A22B is the stronger model overall, scoring 43.5 to 39.2 on the Noometry Index.
Last verified . 26 shared benchmarks.
Alibaba (Qwen)Summary
- They share 26 benchmarks with published results for both. Qwen3 235B-A22B scores higher in 8 categories and Qwen3 32B in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3 235B-A22B leads 50.4 to 39.7.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 86.7% for Qwen3 235B-A22B and 66.9% for Qwen3 32B.
- Both cost about the same: $0.70 input and $2.80 output per million tokens.
Side by side
| Qwen3 235B-A22B | Qwen3 32B | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 43.5 | 39.2 |
| Released | 2025-04 | 2025-04 |
| Weights | Open | Open |
| Context window | 131K | 131K |
| Max output | 16K | 16K |
| Input $ / M tokens | $0.70 | $0.70 |
| Output $ / M tokens | $2.80 | $2.80 |
| Results tracked | 49 | 26 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3 235B-A22B leads
Qwen3 235B-A22B: 44.3 (#75), Qwen3 32B: 37.7 (#190)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| Aider Polyglot | 59.6% | 40% |
| SciCode | 42.4% | 35.4% |
| LMArena Coding | 1445 | 1358 |
| WeirdML | 41% | — |
Agentic & Tool Use Qwen3 235B-A22B leads
Qwen3 235B-A22B: 33.9 (#51), Qwen3 32B: 32.6 (#62)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 52.1% | 48.7% |
| Vending-Bench 2 | -11.34 | — |
Reasoning Qwen3 32B leads
Qwen3 235B-A22B: 15.7 (#311), Qwen3 32B: 20.2 (#241)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| Kagi LLM Benchmark | 69.4% | 54.9% |
| CritPt | 0% | 0.3% |
| Chess Puzzles | 12% | 5% |
| LMArena Hard Prompts | 1433 | 1334 |
| DTBench | 80.3% | 67.5% |
| LMCA | 29.3% | 17.3% |
| Epoch Capabilities Index | 143.85 | 138.51 |
| ARC-AGI-2 | 1.3% | — |
| SimpleBench | 31% | — |
| ARC-AGI-1 | 11% | — |
| Mystery Game Puzzles | 9% | — |
| ForecastBench | 59.7 | — |
Math Qwen3 235B-A22B leads
Qwen3 235B-A22B: 50.4 (#57), Qwen3 32B: 39.7 (#99)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 86.7% | 66.9% |
| LMArena Math | 1432 | 1399 |
| Omni-MATH | 71.8% | — |
| MATH Level 5 | 68.9% | — |
| FrontierMath (Feb 2025 set) | 8.5% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge Qwen3 235B-A22B leads
Qwen3 235B-A22B: 49.6 (#73), Qwen3 32B: 40.0 (#125)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| GPQA Diamond | 80.1% | 65.7% |
| Vectara Hallucination Rate | 9.3% | 5.9% |
| LMArena Expert | 1463 | 1362 |
| SimpleQA Verified | 40.4% | — |
| MMLU-Pro | 84.4% | — |
| Confabulations | 15.6% | — |
| GPQA (HELM) | 72.7% | — |
Multilingual Qwen3 235B-A22B leads
Qwen3 235B-A22B: 52.3 (#89), Qwen3 32B: 45.6 (#167)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| LMArena Non-English | 1409 | 1317 |
| LMArena Chinese | 1481 | 1357 |
| LMArena German | 1433 | 1341 |
| LMArena Russian | 1411 | 1311 |
| LMArena French | 1445 | — |
| LMArena Japanese | 1399 | — |
| LMArena Korean | 1391 | — |
| LMArena Spanish | 1430 | — |
Instruction Following Qwen3 235B-A22B leads
Qwen3 235B-A22B: 72.6 (#136), Qwen3 32B: 68.9 (#179)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| LMArena Instruction Following | 1408 | 1305 |
| IFEval | 83.5% | — |
Long Context Qwen3 235B-A22B leads
Qwen3 235B-A22B: 46.1 (#26), Qwen3 32B: 43.8 (#87)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| Fiction.LiveBench | 75% | 74.2% |
| LMArena Longer Query | 1426 | 1327 |
Writing & Preference Qwen3 235B-A22B leads
Qwen3 235B-A22B: 59.6 (#108), Qwen3 32B: 52.9 (#163)
| Benchmark | Qwen3 235B-A22B | Qwen3 32B |
|---|---|---|
| LMArena Text | 1419 | 1340 |
| LMArena Creative Writing | 1384 | 1297 |
| LMArena Multi-Turn | 1432 | 1331 |
| Short-Story Creative Writing | 83% | — |
| EQ-Bench Creative Writing | 1366 | — |
| WildBench | 86.6% | — |
Frequently asked questions
Is Qwen3 235B-A22B better than Qwen3 32B?
Qwen3 235B-A22B is the stronger model overall, scoring 43.5 to 39.2 on the Noometry Index.
Which is cheaper, Qwen3 235B-A22B or Qwen3 32B?
Qwen3 32B is cheaper. It lists at $0.70 per million input tokens and $2.80 per million output tokens; Qwen3 235B-A22B lists at $0.70 and $2.80.
Is Qwen3 235B-A22B or Qwen3 32B better for coding?
Qwen3 235B-A22B scores higher on coding benchmarks: 44.3 versus 37.7 in the Noometry coding category.
Which has the bigger context window?
Both accept 131K tokens.
How many benchmarks do Qwen3 235B-A22B and Qwen3 32B share?
26 benchmarks have published results for both models. Qwen3 235B-A22B has 49 scored results on Noometry and Qwen3 32B has 26.