by AlibabaRank #47Score 79.1
The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. It delivers...
| Signal | Normalized | Weight | Contribution | Freshness |
|---|---|---|---|---|
Capabilities capability | 83.3 | 20% | 16.7 | 2026-06-24T09:59:39.752Z |
Benchmarks benchmark | 80.3 | 30% | 24.1 | 2026-06-24T09:59:39.752Z |
Pricing pricing_tier | 97.6 | 15% | 14.6 | 2026-06-24T09:59:39.752Z |
Context Window context_window | 77.3 | 10% | 7.7 | 2026-06-24T09:59:39.752Z |
Recency recency | 100.0 | 15% | 15.0 | 2026-06-24T09:59:39.752Z |
Output Capacity output_capacity | 20.0 | 10% | 2.0 | 2026-06-24T09:59:39.752Z |
| Capability | Supported |
|---|---|
| Vision | Yes |
| Reasoning | Yes |
| JSON Mode | Yes |
| Streaming | Yes |
| Function Calling | Yes |
| Web Search | No |