Model comparison
Qwen3.5 27B vs Qwen3.8 Max
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 41.9 on the Noometry Index. Qwen3.5 27B costs 3.6× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
Alibaba (Qwen)Summary
- They share 22 benchmarks with published results for both. Qwen3.5 27B scores higher in 1 category and Qwen3.8 Max in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3.8 Max leads 73.2 to 38.8.
- The biggest single-benchmark swing is NYT Connections (extended): 47.9% for Qwen3.5 27B and 88.3% for Qwen3.8 Max.
- Qwen3.5 27B is cheaper at $0.30 / $2.40 per million input/output tokens, against $2 / $6 for Qwen3.8 Max.
- Qwen3.8 Max accepts more context: 1M tokens versus 262K.
- Qwen3.5 27B has downloadable open weights; the other is API-only.
Side by side
| Qwen3.5 27B | Qwen3.8 Max | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 41.9 | 56.8 |
| Released | 2026-02-23 | 2026-08-02 |
| Weights | Open | Proprietary |
| Context window | 262K | 1M |
| Max output | 66K | 131K |
| Input $ / M tokens | $0.30 | $2 |
| Output $ / M tokens | $2.40 | $6 |
| Results tracked | 28 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 Max leads
Qwen3.5 27B: 38.9 (#168), Qwen3.8 Max: 53.5 (#29)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena WebDev | 1358 | 1674 |
| LMArena Coding | 1427 | 1502 |
| DeepSWE | — | 57.5% |
| FrontierSWE | — | 17.8% |
| SciCode | — | 53.2% |
| WeirdML | 39.5% | — |
| ALE-Bench | 349.45 | — |
Agentic & Tool Use Not comparable
Qwen3.5 27B: —, Qwen3.8 Max: 45.4 (#14)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| APEX-Agents | — | 63.3% |
| τ²-bench Banking | — | 55.1% |
| GDP.pdf | — | 23.2% |
| Vending-Bench 2 | 201.98 | — |
Reasoning Qwen3.8 Max leads
Qwen3.5 27B: 27.5 (#117), Qwen3.8 Max: 54.4 (#26)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| NYT Connections (extended) | 47.9% | 88.3% |
| LMArena Hard Prompts | 1414 | 1496 |
| DTBench | 82.4% | 92% |
| LMCA | 34% | 46.2% |
| CritPt | — | 20% |
| Chess Puzzles | — | 40% |
| Thematic Generalization | 45.5% | — |
| Mystery Game Puzzles | — | 38% |
| Epoch Capabilities Index | — | 156.41 |
Math Qwen3.8 Max leads
Qwen3.5 27B: 38.8 (#127), Qwen3.8 Max: 73.2 (#20)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena Math | 1429 | 1499 |
| FrontierMath (Tiers 1-3) | — | 74.7% |
| FrontierMath Tier 4 | — | 46.3% |
| MathArena Final-Answer Competitions | 56.7% | — |
| OTIS Mock AIME 2024-2025 | — | 100% |
| ProofBench | — | 58% |
Knowledge Qwen3.8 Max leads
Qwen3.5 27B: 38.0 (#150), Qwen3.8 Max: 61.7 (#27)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena Expert | 1428 | 1507 |
| GPQA Diamond | — | 92.7% |
| SimpleQA Verified | — | 47.3% |
| Vectara Hallucination Rate | 12.1% | — |
Multimodal Qwen3.5 27B leads
Qwen3.5 27B: 39.4 (#59), Qwen3.8 Max: 37.2 (#75)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | 1241 | 1314 |
| Furniture Assembly | — | 20% |
Multilingual Qwen3.8 Max leads
Qwen3.5 27B: 50.8 (#115), Qwen3.8 Max: 56.7 (#18)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1390 | 1472 |
| LMArena Chinese | 1478 | 1538 |
| LMArena French | 1410 | 1503 |
| LMArena German | 1393 | 1483 |
| LMArena Japanese | 1345 | 1467 |
| LMArena Korean | 1358 | 1461 |
| LMArena Russian | 1390 | 1481 |
| LMArena Spanish | 1407 | 1492 |
Instruction Following Qwen3.8 Max leads
Qwen3.5 27B: 73.5 (#119), Qwen3.8 Max: 77.6 (#17)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1393 | 1479 |
Long Context Qwen3.8 Max leads
Qwen3.5 27B: 43.1 (#106), Qwen3.8 Max: 45.6 (#31)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1413 | 1489 |
Writing & Preference Qwen3.8 Max leads
Qwen3.5 27B: 59.3 (#111), Qwen3.8 Max: 67.1 (#30)
| Benchmark | Qwen3.5 27B | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1409 | 1483 |
| LMArena Creative Writing | 1362 | 1479 |
| LMArena Multi-Turn | 1410 | 1489 |
Frequently asked questions
Is Qwen3.5 27B better than Qwen3.8 Max?
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 41.9 on the Noometry Index. Qwen3.5 27B costs 3.6× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Which is cheaper, Qwen3.5 27B or Qwen3.8 Max?
Qwen3.5 27B is cheaper. It lists at $0.30 per million input tokens and $2.40 per million output tokens; Qwen3.8 Max lists at $2 and $6.
Is Qwen3.5 27B or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 38.9 in the Noometry coding category.
Which has the bigger context window?
Qwen3.8 Max does, with 1M tokens against 262K.
How many benchmarks do Qwen3.5 27B and Qwen3.8 Max share?
22 benchmarks have published results for both models. Qwen3.5 27B has 28 scored results on Noometry and Qwen3.8 Max has 39.