Qwen 3 — local AI model by Alibaba Cloud
Written by Jakub Rusinowski · Last updated
Alibaba's model series with a hybrid thinking mode. Qwen 3 models can toggle between fast response mode and deep Chain-of-Thought reasoning on demand — type /think for hard problems, /no_think for fast answers. Available in dense and MoE variants. Previous generation — superseded by Qwen 3.5 and Qwen 3.6, which extend context length and improve reasoning efficiency. Still widely used and supported.
Variants
The smallest Qwen 3 variant needs about 6 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 8B → 8B | ~5.8 GB | 128,000 | ollama run qwen3:8b | |
| Qwen 3 14B → 14B | ~9.7 GB | 128,000 | ollama run qwen3:14b | |
| Qwen 3 32B → 32B | ~20.6 GB | 128,000 | ollama run qwen3:32b | |
| Qwen 3 30B-A3B (MoE) → 30B (3B active) | ~19.2 GB | 128,000 | ollama run qwen3:30b-a3b | |
| Qwen 3 235B-A22B (MoE) → 235B (22B active) | ~142.7 GB | 128,000 | ollama run qwen3:235b-a22b |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Qwen 3 locally
Install Ollama, then pull the tag.
ollama run qwen3:8bPick 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 8B, Qwen 3 14B, Qwen 3 32B, Qwen 3 30B-A3B (MoE), Qwen 3 235B-A22B (MoE)Recommended GPU
The cheapest catalogued GPU that runs Qwen 3 locally (min 6 GB VRAM) is the Intel Arc B570 (10 GB).
Can I run Qwen 3 on my GPU?
- Qwen 3 on AMD Radeon RX 7800 XT
- Qwen 3 on AMD Radeon RX 7900 XT
- Qwen 3 on AMD Radeon RX 7900 XTX
- Qwen 3 on AMD Radeon RX 9060 XT 16GB
- Qwen 3 on AMD Radeon RX 9060 XT 8GB
- Qwen 3 on AMD Radeon RX 9070
- Qwen 3 on AMD Radeon RX 9070 XT
- Qwen 3 on Apple M1
- Qwen 3 on Apple M1 Pro
- Qwen 3 on Apple M2 Pro
- Qwen 3 on Apple M2 Ultra
- Qwen 3 on Apple M3 Pro
- Qwen 3 on Apple M4
- Qwen 3 on Apple M5
- Qwen 3 on Intel Arc B570
- Qwen 3 on Intel Arc B580
- Qwen 3 on NVIDIA GeForce RTX 3060 (12GB)
- Qwen 3 on NVIDIA GeForce RTX 3070
- Qwen 3 on NVIDIA GeForce RTX 3070 Ti
- Qwen 3 on NVIDIA GeForce RTX 3080 (10GB)
- Qwen 3 on NVIDIA GeForce RTX 3090
- Qwen 3 on NVIDIA GeForce RTX 4060
- Qwen 3 on NVIDIA GeForce RTX 4060 Ti 16GB
- Qwen 3 on NVIDIA GeForce RTX 4070
- Qwen 3 on NVIDIA GeForce RTX 4070 Super
- Qwen 3 on NVIDIA GeForce RTX 4070 Ti
- Qwen 3 on NVIDIA GeForce RTX 4070 Ti Super
- Qwen 3 on NVIDIA GeForce RTX 4080
- Qwen 3 on NVIDIA GeForce RTX 4080 Super
- Qwen 3 on NVIDIA GeForce RTX 4090
- Qwen 3 on NVIDIA GeForce RTX 5060
- Qwen 3 on NVIDIA GeForce RTX 5060 Ti 16GB
- Qwen 3 on NVIDIA GeForce RTX 5060 Ti 8GB
- Qwen 3 on NVIDIA GeForce RTX 5070
- Qwen 3 on NVIDIA GeForce RTX 5070 Ti
- Qwen 3 on NVIDIA GeForce RTX 5080
- Qwen 3 on NVIDIA GeForce RTX 5090
- Qwen 3 on NVIDIA L40S
- Qwen 3 on NVIDIA RTX 6000 Ada Generation
Qwen 3 — frequently asked questions
How much VRAM does Qwen 3 need?
Qwen 3 needs about 6 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Qwen 3 8B (6 GB, Q4_K_M); Qwen 3 14B (10 GB, Q4_K_M); Qwen 3 32B (21 GB, Q4_K_M); Qwen 3 30B-A3B (MoE) (19 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Qwen 3 on an RTX 4090 (24 GB)?
Yes — Qwen 3 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?
Q4_K_M is the best balance of quality and VRAM for Qwen 3 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 with Ollama?
Install Ollama, then run: ollama run qwen3:8b. This downloads Qwen 3 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.