QwQ

What is QwQ?

QwQ is Alibaba's open-weights language model. Released in 2025-03, it ships in 1 size (32B) and is licensed Apache 2.0. The most popular use case is QwQ reasoning model on a single 24GB GPU.

VRAM and hardware

The 32B size fits in 20 GB of VRAM.

Size	Min VRAM (Q4_K_M)	Recommended VRAM	Tokens/sec on 3090
32B	20 GB	30 GB	~17 tok/s

How to run QwQ locally

Option 1: Ollama (simplest)

ollama pull qwq
ollama run qwq

Option 2: Mullama (production)

mullama pull qwq
mullama run qwq

Option 3: llama.cpp (CLI)

# Download a GGUF from Hugging Face (search "qwq gguf")
./llama-cli -m qwq.Q4_K_M.gguf -p "Hello, AI!"

Option 4: Python with Mullama or llama-cpp-python

from mullama import Model, Context
model = Model.load("qwq.Q4_K_M.gguf", n_gpu_layers=99)
ctx = Context(model, n_ctx=4096)
print(ctx.generate("Hello, AI!", 256))

What you can build with QwQ

• reasoning
• math
• logic

Hardware it fits on

• Apple Silicon (8GB): 32B in MLX
• Apple Silicon (16GB): 32B in MLX
• Apple Silicon (32GB): 32B in MLX or GGUF
• Apple Silicon (64GB+): 32B in MLX
• RTX 3090 (24GB): 32B at Q4_K_M
• RTX 4090 (24GB): same as 3090, ~30% faster
• RTX 5090 (32GB): 32B at Q8_0
• 2× RTX 6000 Ada (96GB total): 32B at Q4_K_M

Related models in the QwQ family

This is the only QwQ-family model in the catalog right now.

Other models you might consider

• Llama 3.1 (Meta) — 8B, general-purpose chat, agentic workflows, code
• Llama 3.2 (Meta) — 1B, low-end hardware, mobile, edge deployment
• Llama 3.2 Vision (Meta) — 11B, vision + chat on desktop GPUs
• Llama 3.3 (Meta) — 70B, best-in-class 70B for desktops
• Llama 4 (Meta) — Scout 17B (109B MoE), vision + long context (10M tokens)
• Qwen 2.5 (Alibaba) — 0.5B, strongest 7B-32B on consumer hardware