Qwen3.5 9B by Alibaba: dense open-weight small-series VLM with Gated DeltaNet hybrid layers, early fusion multimodal pretraining, MTP, and 201-language support.
Capabilities, design details, and architectural traits
Qwen3.5 9B is the top-of-range model in the officially designated Small subfamily (0.8B, 2B, 4B, 9B) of the Qwen3.5 family. Like the 27B, it is a dense model with a built-in vision encoder - every parameter is active during inference and image/video inputs are handled natively without a separately attached adapter.
| Trait | Detail |
|---|---|
| Gated DeltaNet hybrid layout | 32 layers arranged in 8 cycles: each cycle contains 3x (Gated DeltaNet -> FFN) followed by 1x (Gated Attention -> FFN), interleaving linear and standard attention at a fixed 3:1 ratio |
| Early fusion multimodal pretraining | Text, image, and video tokens are interleaved and trained jointly from pretraining - not via post-hoc visual adapters |
| Multi-Token Prediction (MTP) | Trained with multi-step MTP, documented explicitly in the model card |
| 201 languages | Multilingual coverage across 201 languages and dialects |
| Context window | 262,144 tokens natively, extensible to 1,010,000 tokens via YaRN |
| License | Apache 2.0 open-weight |
The official model card deploys Qwen3.5 9B with --tp-size 1 in both SGLang and vLLM examples, confirming single-GPU operation. The Qwen3.5 Small classification is an official series designation used in Qwen's own documentation to group the 0.8B-9B variants as a distinct subfamily from the Medium models (27B, 35B-A3B, 122B-A10B, 397B-A17B).
Independent evaluations · Artificial Analysis
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Alibaba | Anthropic | Anthropic |
| Release Date | March 2, 2026 | July 24, 2026 | June 9, 2026 |
| Knowledge Cutoff | - | May 2026 | - |
| Context & Limits | |||
| Context Window | 262K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.17 Best Input Pricing | $5 | $10 |
| Output Pricing | $0.25 Best Output Pricing | $25 | $50 |
| Modalities | |||
| Inputs | textimagevideo | textimage | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 20.6 | 63.1 Best Intelligence Index | 62.1 |
| Coding Index | 23.5 | 78.0 Best Coding Index | 76.5 |
| Agentic Index | - | 59.2 Best Agentic Index | 56.6 |
Graduate-level reasoning and expert Q&A evaluation.
Extremely difficult logical reasoning and knowledge.
Logical reasoning over long context windows.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
