Model comparison
Qwen3 32B vs Qwen3.8 Max
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 39.2 on the Noometry Index. Qwen3 32B costs 2.4× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Last verified . 21 shared benchmarks.
Alibaba (Qwen)Summary
- They share 21 benchmarks with published results for both. Qwen3 32B scores higher in 0 categories and Qwen3.8 Max in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Qwen3.8 Max leads 54.4 to 20.2.
- The biggest single-benchmark swing is Chess Puzzles: 5% for Qwen3 32B and 40% for Qwen3.8 Max.
- Qwen3 32B is cheaper at $0.70 / $2.80 per million input/output tokens, against $2 / $6 for Qwen3.8 Max.
- Qwen3.8 Max accepts more context: 1M tokens versus 131K.
- Qwen3 32B has downloadable open weights; the other is API-only.
Side by side
| Qwen3 32B | Qwen3.8 Max | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 39.2 | 56.8 |
| Released | 2025-04 | 2026-08-02 |
| Weights | Open | Proprietary |
| Context window | 131K | 1M |
| Max output | 16K | 131K |
| Input $ / M tokens | $0.70 | $2 |
| Output $ / M tokens | $2.80 | $6 |
| Results tracked | 26 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 Max leads
Qwen3 32B: 37.7 (#190), Qwen3.8 Max: 53.5 (#29)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| SciCode | 35.4% | 53.2% |
| LMArena Coding | 1358 | 1502 |
| DeepSWE | — | 57.5% |
| Aider Polyglot | 40% | — |
| LMArena WebDev | — | 1674 |
| FrontierSWE | — | 17.8% |
Agentic & Tool Use Qwen3.8 Max leads
Qwen3 32B: 32.6 (#62), Qwen3.8 Max: 45.4 (#14)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| APEX-Agents | — | 63.3% |
| Berkeley Function Calling Leaderboard | 48.7% | — |
| τ²-bench Banking | — | 55.1% |
| GDP.pdf | — | 23.2% |
Reasoning Qwen3.8 Max leads
Qwen3 32B: 20.2 (#241), Qwen3.8 Max: 54.4 (#26)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| CritPt | 0.3% | 20% |
| Chess Puzzles | 5% | 40% |
| LMArena Hard Prompts | 1334 | 1496 |
| DTBench | 67.5% | 92% |
| LMCA | 17.3% | 46.2% |
| Epoch Capabilities Index | 138.51 | 156.41 |
| Kagi LLM Benchmark | 54.9% | — |
| NYT Connections (extended) | — | 88.3% |
| Mystery Game Puzzles | — | 38% |
Math Qwen3.8 Max leads
Qwen3 32B: 39.7 (#99), Qwen3.8 Max: 73.2 (#20)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.9% | 100% |
| LMArena Math | 1399 | 1499 |
| FrontierMath (Tiers 1-3) | — | 74.7% |
| FrontierMath Tier 4 | — | 46.3% |
| ProofBench | — | 58% |
Knowledge Qwen3.8 Max leads
Qwen3 32B: 40.0 (#125), Qwen3.8 Max: 61.7 (#27)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| GPQA Diamond | 65.7% | 92.7% |
| LMArena Expert | 1362 | 1507 |
| SimpleQA Verified | — | 47.3% |
| Vectara Hallucination Rate | 5.9% | — |
Multimodal Not comparable
Qwen3 32B: —, Qwen3.8 Max: 37.2 (#75)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | — | 1314 |
| Furniture Assembly | — | 20% |
Multilingual Qwen3.8 Max leads
Qwen3 32B: 45.6 (#167), Qwen3.8 Max: 56.7 (#18)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1317 | 1472 |
| LMArena Chinese | 1357 | 1538 |
| LMArena German | 1341 | 1483 |
| LMArena Russian | 1311 | 1481 |
| LMArena French | — | 1503 |
| LMArena Japanese | — | 1467 |
| LMArena Korean | — | 1461 |
| LMArena Spanish | — | 1492 |
Instruction Following Qwen3.8 Max leads
Qwen3 32B: 68.9 (#179), Qwen3.8 Max: 77.6 (#17)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1305 | 1479 |
Long Context Qwen3.8 Max leads
Qwen3 32B: 43.8 (#87), Qwen3.8 Max: 45.6 (#31)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1327 | 1489 |
| Fiction.LiveBench | 74.2% | — |
Writing & Preference Qwen3.8 Max leads
Qwen3 32B: 52.9 (#163), Qwen3.8 Max: 67.1 (#30)
| Benchmark | Qwen3 32B | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1340 | 1483 |
| LMArena Creative Writing | 1297 | 1479 |
| LMArena Multi-Turn | 1331 | 1489 |
Frequently asked questions
Is Qwen3 32B better than Qwen3.8 Max?
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 39.2 on the Noometry Index. Qwen3 32B costs 2.4× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Which is cheaper, Qwen3 32B or Qwen3.8 Max?
Qwen3 32B is cheaper. It lists at $0.70 per million input tokens and $2.80 per million output tokens; Qwen3.8 Max lists at $2 and $6.
Is Qwen3 32B or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 37.7 in the Noometry coding category.
Which has the bigger context window?
Qwen3.8 Max does, with 1M tokens against 131K.
How many benchmarks do Qwen3 32B and Qwen3.8 Max share?
21 benchmarks have published results for both models. Qwen3 32B has 26 scored results on Noometry and Qwen3.8 Max has 39.