Qwen3.5-9B
Qwen3.5-9B by Alibaba Qwen (9B parameters). At Q4_K_M with an 8K context it needs about 6.1 GiB of memory. A 9B model that reads images and video, with a 256K context; only 8 of its 32 layers keep a KV cache, so long contexts use little memory. It fits a 12 GB card at 8-bit.
Last updated: 2026-10-13
Model facts
Run it
ollama run qwen3.5:9b
Memory needed
Weights + KV cache + runtime overhead, in GiB. Lower quantization = less memory, slightly lower quality.
| Quantization | 4K context | 8K context | 32K context |
|---|---|---|---|
| Q4_K_M | 6.0 GiB | 6.1 GiB | 6.8 GiB |
| Q5_K_M | 6.9 GiB | 7.0 GiB | 7.8 GiB |
| Q6_K | 7.9 GiB | 8.0 GiB | 8.7 GiB |
| Q8_0 | 10.0 GiB | 10.1 GiB | 10.9 GiB |
| FP16 | 18.2 GiB | 18.4 GiB | 19.1 GiB |
Which hardware runs it
Best quantization that fits each device with a 8K context. Estimates, not benchmarks. See the methodology.
Frequently asked questions
How much VRAM does Qwen3.5-9B need?
With an 8K context, about 6.1 GiB at Q4_K_M and 10.1 GiB at Q8_0 (weights + KV cache + runtime overhead). Longer contexts need more memory for the KV cache.
Can Qwen3.5-9B run on an GeForce RTX 4090?
Yes. Qwen3.5-9B fits on the GeForce RTX 4090 (24 GB) at FP16 with an 8K context, using about 18.4 GiB, at an estimated 39 tokens/s.
Can I run Qwen3.5-9B on a Mac?
Yes. Qwen3.5-9B fits on the M3 Ultra (96 GB) at FP16 with an 8K context, using about 18.4 GiB, at an estimated 30 tokens/s.
Can Qwen3.5-9B be used commercially?
Qwen3.5-9B is released under Apache-2.0. The licence allows commercial use (read its conditions).
Sources
- https://github.com/QwenLM/Qwen3.8/blob/main/README.md
- https://huggingface.co/Qwen/Qwen3.5-9B/blob/main/LICENSE
- https://huggingface.co/Qwen/Qwen3.5-9B
- https://huggingface.co/Qwen/Qwen3.5-9B/resolve/main/config.json
- https://huggingface.co/api/models/Qwen/Qwen3.5-9B?expand[]=safetensors
- https://ollama.com/library/qwen3.5/tags