Qwen 3.5 — local AI model by Alibaba Cloud
Written by Jakub Rusinowski · Last updated
Alibaba's next-generation flagship family, released in waves from Feb 16 to Mar 2, 2026 (397B-A17B first, then the 122B/35B-A3B/27B tier on Feb 24, then the 0.8B–9B 'Small' series on Mar 2). Qwen 3.5 introduces a hybrid Gated DeltaNet + MoE architecture delivering frontier performance at a fraction of the compute, with native multimodal vision, ~256K context (extensible to 1M on the larger tiers), 201 languages, and Apache 2.0 license. Note: the 397B-A17B flagship is only available as a cloud-hosted Ollama tag (`:cloud`), not a true local download — everything 0.8B–122B-A10B has standard local Ollama tags.
Variants
The smallest Qwen 3.5 variant needs about 1 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Qwen 3.5 0.8B → 0.8B | ~1.3 GB | 256,000 | ollama run qwen3.5:0.8b | |
| Qwen 3.5 2B → 2B | ~2 GB | 256,000 | ollama run qwen3.5:2b | |
| Qwen 3.5 4B → 4B | ~3.2 GB | 256,000 | ollama run qwen3.5:4b | |
| Qwen 3.5 9B → 9B | ~6.2 GB | 256,000 | ollama run qwen3.5:9b | |
| Qwen 3.5 27B → 27B | ~17.1 GB | 262,144 | ollama run qwen3.5:27b | |
| Qwen 3.5 35B-A3B → 35B (3B active) | ~21.9 GB | 262,144 | ollama run qwen3.5:35b-a3b | |
| Qwen 3.5 122B-A10B → 122B (10B active) | ~74.5 GB | 262,144 | ollama run qwen3.5:122b | |
| Qwen 3.5 397B-A17B → 397B (17B active) | ~240.5 GB | 262,144 | cloud-hosted tag |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Qwen 3.5 locally
Install Ollama, then pull the tag.
ollama run qwen3.5:0.8bPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Qwen 3.5 0.8B, Qwen 3.5 2B, Qwen 3.5 4B, Qwen 3.5 9B, Qwen 3.5 27B, Qwen 3.5 35B-A3B, Qwen 3.5 122B-A10B, Qwen 3.5 397B-A17BRecommended GPU
The cheapest catalogued GPU that runs Qwen 3.5 locally (min 1 GB VRAM) is the Intel Arc B570 (10 GB).
Can I run Qwen 3.5 on my GPU?
- Qwen 3.5 on AMD Radeon RX 7900 XT
- Qwen 3.5 on AMD Radeon RX 7900 XTX
- Qwen 3.5 on AMD Radeon RX 9060 XT 8GB
- Qwen 3.5 on AMD Ryzen AI Max+ 395
- Qwen 3.5 on Apple M1 Pro
- Qwen 3.5 on Apple M1 Ultra
- Qwen 3.5 on Apple M2
- Qwen 3.5 on Apple M2 Pro
- Qwen 3.5 on Apple M2 Ultra
- Qwen 3.5 on Apple M3
- Qwen 3.5 on Apple M3 Pro
- Qwen 3.5 on Apple M4
- Qwen 3.5 on Apple M4 Max
- Qwen 3.5 on Apple M4 Pro
- Qwen 3.5 on Apple M5
- Qwen 3.5 on Apple M5 Max
- Qwen 3.5 on Intel Arc B570
- Qwen 3.5 on Intel Arc B580
- Qwen 3.5 on NVIDIA A100 80GB (PCIe)
- Qwen 3.5 on NVIDIA DGX Spark
- Qwen 3.5 on NVIDIA GeForce RTX 3060 (12GB)
- Qwen 3.5 on NVIDIA GeForce RTX 3070
- Qwen 3.5 on NVIDIA GeForce RTX 3070 Ti
- Qwen 3.5 on NVIDIA GeForce RTX 3080 (10GB)
- Qwen 3.5 on NVIDIA GeForce RTX 3090
- Qwen 3.5 on NVIDIA GeForce RTX 4060
- Qwen 3.5 on NVIDIA GeForce RTX 4070
- Qwen 3.5 on NVIDIA GeForce RTX 4070 Super
- Qwen 3.5 on NVIDIA GeForce RTX 4070 Ti
- Qwen 3.5 on NVIDIA GeForce RTX 4090
- Qwen 3.5 on NVIDIA GeForce RTX 5060
- Qwen 3.5 on NVIDIA GeForce RTX 5060 Ti 8GB
- Qwen 3.5 on NVIDIA GeForce RTX 5070
- Qwen 3.5 on NVIDIA GeForce RTX 5090
- Qwen 3.5 on NVIDIA H100 80GB (PCIe)
- Qwen 3.5 on NVIDIA L40S
- Qwen 3.5 on NVIDIA RTX 6000 Ada Generation
- Qwen 3.5 on NVIDIA RTX PRO 6000 Blackwell
Qwen 3.5 — frequently asked questions
How much VRAM does Qwen 3.5 need?
Qwen 3.5 needs about 1 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Qwen 3.5 0.8B (1 GB, Q4_K_M); Qwen 3.5 2B (2 GB, Q4_K_M); Qwen 3.5 4B (3 GB, Q4_K_M); Qwen 3.5 9B (6 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Qwen 3.5 on an RTX 4090 (24 GB)?
Yes — Qwen 3.5 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Qwen 3.5?
Q4_K_M is the best balance of quality and VRAM for Qwen 3.5 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Qwen 3.5 with Ollama?
Install Ollama, then run: ollama run qwen3.5:0.8b. This downloads Qwen 3.5 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.