Model comparison
Qwen3.5 397B-A17B vs Qwen3.8 Max
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 46.0 on the Noometry Index. Qwen3.5 397B-A17B costs 2.2× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Last verified . 30 shared benchmarks.
Alibaba (Qwen)Summary
- They share 30 benchmarks with published results for both. Qwen3.5 397B-A17B scores higher in 1 category and Qwen3.8 Max in 9 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3.8 Max leads 73.2 to 46.1.
- The biggest single-benchmark swing is τ²-bench Banking: 9.8% for Qwen3.5 397B-A17B and 55.1% for Qwen3.8 Max.
- Qwen3.5 397B-A17B is cheaper at $0.60 / $3.60 per million input/output tokens, against $2 / $6 for Qwen3.8 Max.
- Qwen3.8 Max accepts more context: 1M tokens versus 262K.
- Qwen3.5 397B-A17B has downloadable open weights; the other is API-only.
Side by side
| Qwen3.5 397B-A17B | Qwen3.8 Max | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 46.0 | 56.8 |
| Released | 2026-02-01 | 2026-08-02 |
| Weights | Open | Proprietary |
| Context window | 262K | 1M |
| Max output | 66K | 131K |
| Input $ / M tokens | $0.60 | $2 |
| Output $ / M tokens | $3.60 | $6 |
| Results tracked | 36 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 Max leads
Qwen3.5 397B-A17B: 42.0 (#114), Qwen3.8 Max: 53.5 (#29)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| LMArena WebDev | 1400 | 1674 |
| LMArena Coding | 1465 | 1502 |
| DeepSWE | — | 57.5% |
| FrontierSWE | — | 17.8% |
| SciCode | — | 53.2% |
Agentic & Tool Use Qwen3.8 Max leads
Qwen3.5 397B-A17B: 33.3 (#53), Qwen3.8 Max: 45.4 (#14)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| APEX-Agents | 24.9% | 63.3% |
| τ²-bench Banking | 9.8% | 55.1% |
| τ²-bench Airline | 81.5% | — |
| τ²-bench Retail | 84.4% | — |
| τ²-bench Telecom | 97.8% | — |
| GDP.pdf | — | 23.2% |
Reasoning Qwen3.8 Max leads
Qwen3.5 397B-A17B: 34.5 (#70), Qwen3.8 Max: 54.4 (#26)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| NYT Connections (extended) | 58.9% | 88.3% |
| Chess Puzzles | 13% | 40% |
| LMArena Hard Prompts | 1448 | 1496 |
| Mystery Game Puzzles | 18% | 38% |
| DTBench | 87.5% | 92% |
| LMCA | 37.9% | 46.2% |
| Epoch Capabilities Index | 146.65 | 156.41 |
| Kagi LLM Benchmark | 73.7% | — |
| CritPt | — | 20% |
| Thematic Generalization | 65.1% | — |
Math Qwen3.8 Max leads
Qwen3.5 397B-A17B: 46.1 (#73), Qwen3.8 Max: 73.2 (#20)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| FrontierMath (Tiers 1-3) | 31.2% | 74.7% |
| OTIS Mock AIME 2024-2025 | 88.9% | 100% |
| LMArena Math | 1454 | 1499 |
| FrontierMath Tier 4 | — | 46.3% |
| ProofBench | — | 58% |
Knowledge Qwen3.8 Max leads
Qwen3.5 397B-A17B: 53.3 (#58), Qwen3.8 Max: 61.7 (#27)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| GPQA Diamond | 86.4% | 92.7% |
| LMArena Expert | 1462 | 1507 |
| SimpleQA Verified | — | 47.3% |
Multimodal Qwen3.5 397B-A17B leads
Qwen3.5 397B-A17B: 40.7 (#44), Qwen3.8 Max: 37.2 (#75)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | 1263 | 1314 |
| Furniture Assembly | — | 20% |
Multilingual Qwen3.8 Max leads
Qwen3.5 397B-A17B: 53.7 (#59), Qwen3.8 Max: 56.7 (#18)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1430 | 1472 |
| LMArena Chinese | 1500 | 1538 |
| LMArena French | 1461 | 1503 |
| LMArena German | 1447 | 1483 |
| LMArena Japanese | 1426 | 1467 |
| LMArena Korean | 1384 | 1461 |
| LMArena Russian | 1429 | 1481 |
| LMArena Spanish | 1441 | 1492 |
Instruction Following Qwen3.8 Max leads
Qwen3.5 397B-A17B: 75.0 (#77), Qwen3.8 Max: 77.6 (#17)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1424 | 1479 |
Long Context Qwen3.8 Max leads
Qwen3.5 397B-A17B: 44.1 (#74), Qwen3.8 Max: 45.6 (#31)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1442 | 1489 |
Writing & Preference Qwen3.8 Max leads
Qwen3.5 397B-A17B: 62.3 (#79), Qwen3.8 Max: 67.1 (#30)
| Benchmark | Qwen3.5 397B-A17B | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1438 | 1483 |
| LMArena Creative Writing | 1401 | 1479 |
| LMArena Multi-Turn | 1446 | 1489 |
| EQ-Bench Creative Writing | 1478 | — |
Frequently asked questions
Is Qwen3.5 397B-A17B better than Qwen3.8 Max?
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 46.0 on the Noometry Index. Qwen3.5 397B-A17B costs 2.2× 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 397B-A17B or Qwen3.8 Max?
Qwen3.5 397B-A17B is cheaper. It lists at $0.60 per million input tokens and $3.60 per million output tokens; Qwen3.8 Max lists at $2 and $6.
Is Qwen3.5 397B-A17B or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 42.0 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 397B-A17B and Qwen3.8 Max share?
30 benchmarks have published results for both models. Qwen3.5 397B-A17B has 36 scored results on Noometry and Qwen3.8 Max has 39.