Head-to-head
Qwen3.8-Max vs Qwen3.5-397B-A17B
A 0.8-point composite gap — but the real story is the dimension mix. Same weights, unpacked: who leads where, and whether the premium is justified.
#3 / 24
Qwen3.5-397B-A17B
Alibaba
The multimodal flagship that behaves like infrastructure
Full dossierThe short answer
The composites differ by 0.8 points, and their strengths do not overlap.
Stronger overall: Qwen3.8-Max. If you only run one workload, read the dimension table instead.
Overlaid radar
Same weights, same scale; overlap means parity
- Qwen3.8-Max95.4
- Qwen3.5-397B-A17B94.6
Dimension deltas
A on the left, B on the right; 100 max per row
Qwen3.8-Max leads 5, Qwen3.5-397B-A17B leads 1, tied 2.
Spec sheet
Hard specs in one table
| Item | Qwen3.8-Max | Qwen3.5-397B-A17B |
|---|---|---|
| Composite | 95.4 | 94.6 |
| Tier | S+ | S+ |
| Visual perception & grounding | 97WIN | 96 |
| Document parsing & Chinese OCR | 97 | 97 |
| Chart & document reasoning | 96WIN | 95 |
| Video & long-horizon reasoning | 93WIN | 92 |
| GUI agents & tool use | 95WIN | 92 |
| Generation & editing | 90 | 90 |
| Chinese context & culture | 95WIN | 94 |
| Ecosystem, cost & availability | 96 | 98WIN |
| Output price | ¥18 / M | ¥12 / M |
| Context | 1M (256K native) | 256K (up to 1M) |
| Parameters | Undisclosed (MoE, trillion-scale total) | 397B total / 17B active (MoE) |
| Licence | Closed commercial API (smaller siblings open) | Smaller sizes open (Apache-2.0); flagship closed |
| Native multimodal | Yes | Yes |
| Evidence grade | B | A |
How to choose
Choose Qwen3.8-Max if…
- Long-document, chart-dense enterprise knowledge bases
- Whole-manual QA that needs 1M context
- Cross-app GUI agents and automation flows
- You cannot accept its limitation: At ¥18/1M output tokens it is one of the three priciest here; high-frequency image analysis gets a heavy bill.
How to choose
Choose Qwen3.5-397B-A17B if…
- Enterprise document hubs and contract review
- Multilingual product data extraction
- On-prem public-sector projects
- You cannot accept its limitation: The flagship is closed; only small and mid sizes are truly self-hostable.
Scenario check
Five scenarios, who holds up
| Scenario | Qwen3.8-Max | Qwen3.5-397B-A17B | Better |
|---|---|---|---|
| Chinese receipt OCR | 97 | 96 | Qwen3.8-Max |
| Chart data extraction | 96 | 95 | Qwen3.8-Max |
| 30-min video QA | 92 | 91 | Qwen3.8-Max |
| Mobile GUI control | 95 | 91 | Qwen3.8-Max |
| Chinese cultural context | 95 | 94 | Qwen3.8-Max |
Comparison FAQ
Qwen3.8-Max vs Qwen3.5-397B-A17B: which is stronger?
Composite: Qwen3.8-Max 95.4, Qwen3.5-397B-A17B 94.6. Qwen3.8-Max leads on Visual perception & grounding, Chart & document reasoning, Video & long-horizon reasoning, GUI agents & tool use, Chinese context & culture; Qwen3.5-397B-A17B leads on Ecosystem, cost & availability. The stronger overall is Qwen3.8-Max.
How much more expensive is Qwen3.8-Max than Qwen3.5-397B-A17B?
Qwen3.8-Max is ¥18/1M output tokens; Qwen3.5-397B-A17B is ¥12/1M. On output price, Qwen3.8-Max is pricier; real cost also depends on image tokens and input share.
Can Qwen3.5-397B-A17B replace Qwen3.8-Max?
It depends on your primary workload. If it sits in Qwen3.5-397B-A17B’s strengths (Enterprise document hubs and contract review, Multilingual product data extraction, On-prem public-sector projects), yes. If you need Long-document, chart-dense enterprise knowledge bases, switching will cost quality.
Personal ranking · Not an official score
Every ranking, score and comparison on this page is an independent personal ranking, not an official score, and represents no vendor, institution or regulator. Scores are a subjective mapping of public evidence under our own weights — change the weights and the order changes. Disagree, or spotted a factual error? Contact the site owner: jizhi2024@foxmail.com — we check each report and log corrections in the changelog.